repo
stringlengths
2
99
file
stringlengths
13
225
code
stringlengths
0
18.3M
file_length
int64
0
18.3M
avg_line_length
float64
0
1.36M
max_line_length
int64
0
4.26M
extension_type
stringclasses
1 value
flair
flair-master/flair/trainers/plugins/functional/__init__.py
0
0
0
py
flair
flair-master/flair/trainers/plugins/functional/anneal_on_plateau.py
import logging import os from flair.trainers.plugins.base import TrainerPlugin, TrainingInterrupt from flair.trainers.plugins.metric_records import MetricRecord from flair.training_utils import AnnealOnPlateau log = logging.getLogger("flair") class AnnealingPlugin(TrainerPlugin): """Plugin for annealing logic in Flair.""" def __init__( self, base_path, min_learning_rate, anneal_factor, patience, initial_extra_patience, anneal_with_restarts, ) -> None: super().__init__() # path to store the model self.base_path = base_path # special annealing modes self.anneal_with_restarts = anneal_with_restarts # determine the min learning rate self.min_learning_rate = min_learning_rate self.anneal_factor = anneal_factor self.patience = patience self.initial_extra_patience = initial_extra_patience def store_learning_rate(self): optimizer = self.trainer.optimizer self.current_learning_rate = [group["lr"] for group in optimizer.param_groups] self.current_momentum = [ group["betas"][0] if "betas" in group else group.get("momentum", 0) for group in optimizer.param_groups ] @TrainerPlugin.hook def after_setup( self, train_with_dev, optimizer, **kw, ): """Initialize different schedulers, including anneal target for AnnealOnPlateau, batch_growth_annealing, loading schedulers. :param train_with_dev: :param optimizer: :param kw: :return: """ # minimize training loss if training with dev data, else maximize dev score anneal_mode = "min" if train_with_dev else "max" # instantiate the scheduler self.scheduler: AnnealOnPlateau = AnnealOnPlateau( factor=self.anneal_factor, patience=self.patience, initial_extra_patience=self.initial_extra_patience, mode=anneal_mode, verbose=False, optimizer=self.trainer.optimizer, ) self.store_learning_rate() @TrainerPlugin.hook def after_evaluation(self, current_model_is_best, validation_scores, **kw): """Scheduler step of AnnealOnPlateau. :param current_model_is_best: :param validation_scores: :param kw: :return: """ reduced_learning_rate: bool = self.scheduler.step(*validation_scores) self.store_learning_rate() bad_epochs = self.scheduler.num_bad_epochs if reduced_learning_rate: bad_epochs = self.patience + 1 log.info( f" - {bad_epochs} epochs without improvement (above 'patience')" f"-> annealing learning_rate to {self.current_learning_rate}" ) else: log.info(f" - {bad_epochs} epochs without improvement") self.trainer.dispatch( "metric_recorded", MetricRecord.scalar(name="bad_epochs", value=bad_epochs, global_step=self.scheduler.last_epoch + 1), ) # stop training if learning rate becomes too small for lr in self.current_learning_rate: if lr < self.min_learning_rate: raise TrainingInterrupt("learning rate too small - quitting training!") # reload last best model if annealing with restarts is enabled if self.anneal_with_restarts and reduced_learning_rate and os.path.exists(self.base_path / "best-model.pt"): log.info("resetting to best model") self.model.load_state_dict(self.model.load(self.base_path / "best-model.pt").state_dict()) def __str__(self) -> str: return ( f"AnnealOnPlateau | " f"patience: '{self.patience}', " f"anneal_factor: '{self.anneal_factor}', " f"min_learning_rate: '{self.min_learning_rate}'" )
3,975
31.859504
132
py
flair
flair-master/tests/test_lemmatizer.py
import torch import flair from flair.data import Sentence from flair.models import Lemmatizer def test_words_to_char_indices(): sentence = Sentence("Hello look what a beautiful day!") lemmatizer = Lemmatizer() # lemmatizer uses standard char dictionary d = lemmatizer.dummy_index e = lemmatizer.end_index s = lemmatizer.start_index string_list = sentence.to_tokenized_string().split() # With end symbol, without start symbol, padding in front target = torch.tensor( [ [d, d, d, d, 55, 5, 15, 15, 12, e], [d, d, d, d, d, 15, 12, 12, 28, e], [d, d, d, d, d, 23, 13, 9, 8, e], [d, d, d, d, d, d, d, d, 9, e], [24, 5, 9, 16, 8, 7, 22, 16, 15, e], [d, d, d, d, d, d, 14, 9, 27, e], [d, d, d, d, d, d, d, d, 76, e], ], dtype=torch.long, ).to(flair.device) out = lemmatizer.words_to_char_indices(string_list, end_symbol=True, start_symbol=False, padding_in_front=True) assert torch.equal(target, out) # Without end symbol, with start symbol, padding in back target = torch.tensor( [ [s, 55, 5, 15, 15, 12, d, d, d, d], [s, 15, 12, 12, 28, d, d, d, d, d], [s, 23, 13, 9, 8, d, d, d, d, d], [s, 9, d, d, d, d, d, d, d, d], [s, 24, 5, 9, 16, 8, 7, 22, 16, 15], [s, 14, 9, 27, d, d, d, d, d, d], [s, 76, d, d, d, d, d, d, d, d], ], dtype=torch.long, ).to(flair.device) out = lemmatizer.words_to_char_indices(string_list, end_symbol=False, start_symbol=True, padding_in_front=False) assert torch.equal(target, out) # Without end symbol, without start symbol, padding in front assert lemmatizer.words_to_char_indices( string_list, end_symbol=False, start_symbol=False, padding_in_front=True ).size() == (7, 9)
1,907
33.690909
116
py
flair
flair-master/tests/test_labels.py
from typing import List from flair.data import Label, Relation, Sentence, Span def test_token_tags(): # example sentence sentence = Sentence("I love Berlin") # set 4 labels for 2 tokens ('love' is tagged twice) sentence[1].add_label("pos", "verb") sentence[1].add_label("sentiment", "positive") sentence[2].add_label("pos", "proper noun") sentence[0].add_label("pos", "pronoun") # check if there are three POS labels with correct text and values labels: List[Label] = sentence.get_labels("pos") assert len(labels) == 3 assert labels[0].data_point.text == "I" assert labels[0].value == "pronoun" assert labels[1].data_point.text == "love" assert labels[1].value == "verb" assert labels[2].data_point.text == "Berlin" assert labels[2].value == "proper noun" # check if there are is one SENTIMENT label with correct text and values labels: List[Label] = sentence.get_labels("sentiment") assert len(labels) == 1 assert labels[0].data_point.text == "love" assert labels[0].value == "positive" # check if all tokens are correctly labeled assert len(sentence) == 3 assert sentence[0].text == "I" assert sentence[1].text == "love" assert sentence[2].text == "Berlin" assert len(sentence[0].get_labels("pos")) == 1 assert len(sentence[1].get_labels("pos")) == 1 assert len(sentence[1].labels) == 2 assert len(sentence[2].get_labels("pos")) == 1 assert sentence[1].get_label("pos").value == "verb" assert sentence[1].get_label("sentiment").value == "positive" # remove the pos label from the last word sentence[2].remove_labels("pos") # there should be 2 POS labels left labels: List[Label] = sentence.get_labels("pos") assert len(labels) == 2 assert len(sentence[0].get_labels("pos")) == 1 assert len(sentence[1].get_labels("pos")) == 1 assert len(sentence[1].labels) == 2 assert len(sentence[2].get_labels("pos")) == 0 # now remove all pos tags sentence.remove_labels("pos") print(sentence[0].get_labels("pos")) assert len(sentence.get_labels("pos")) == 0 assert len(sentence.get_labels("sentiment")) == 1 assert len(sentence.labels) == 1 assert len(sentence[0].get_labels("pos")) == 0 assert len(sentence[1].get_labels("pos")) == 0 assert len(sentence[2].get_labels("pos")) == 0 def test_span_tags(): # set 3 labels for 2 spans (HU is tagged twice) sentence = Sentence("Humboldt Universität zu Berlin is located in Berlin .") sentence[0:4].add_label("ner", "Organization") sentence[0:4].add_label("ner", "University") sentence[7:8].add_label("ner", "City") # check if there are three labels with correct text and values labels: List[Label] = sentence.get_labels("ner") assert len(labels) == 3 assert labels[0].data_point.text == "Humboldt Universität zu Berlin" assert labels[0].value == "Organization" assert labels[1].data_point.text == "Humboldt Universität zu Berlin" assert labels[1].value == "University" assert labels[2].data_point.text == "Berlin" assert labels[2].value == "City" # check if there are two spans with correct text and values spans: List[Span] = sentence.get_spans("ner") assert len(spans) == 2 assert spans[0].text == "Humboldt Universität zu Berlin" assert len(spans[0].get_labels("ner")) == 2 assert spans[1].text == "Berlin" assert spans[1].get_label("ner").value == "City" # now delete the NER tags of "Humboldt-Universität zu Berlin" sentence[0:4].remove_labels("ner") # should be only one NER label left labels: List[Label] = sentence.get_labels("ner") assert len(labels) == 1 assert labels[0].data_point.text == "Berlin" assert labels[0].value == "City" # and only one NER span spans: List[Span] = sentence.get_spans("ner") assert len(spans) == 1 assert spans[0].text == "Berlin" assert spans[0].get_label("ner").value == "City" def test_different_span_tags(): # set 3 labels for 2 spans (HU is tagged twice with different tags) sentence = Sentence("Humboldt Universität zu Berlin is located in Berlin .") sentence[0:4].add_label("ner", "Organization") sentence[0:4].add_label("orgtype", "University") sentence[7:8].add_label("ner", "City") # check if there are three labels with correct text and values labels: List[Label] = sentence.get_labels("ner") assert len(labels) == 2 assert labels[0].data_point.text == "Humboldt Universität zu Berlin" assert labels[0].value == "Organization" assert labels[1].data_point.text == "Berlin" assert labels[1].value == "City" # check if there are two spans with correct text and values spans: List[Span] = sentence.get_spans("ner") assert len(spans) == 2 assert spans[0].text == "Humboldt Universität zu Berlin" assert spans[0].get_label("ner").value == "Organization" assert spans[0].get_label("orgtype").value == "University" assert len(spans[0].get_labels("ner")) == 1 assert spans[1].text == "Berlin" assert spans[1].get_label("ner").value == "City" # now delete the NER tags of "Humboldt-Universität zu Berlin" sentence[0:4].remove_labels("ner") # should be only one NER label left labels: List[Label] = sentence.get_labels("ner") assert len(labels) == 1 assert labels[0].data_point.text == "Berlin" assert labels[0].value == "City" # and only one NER span spans: List[Span] = sentence.get_spans("ner") assert len(spans) == 1 assert spans[0].text == "Berlin" assert spans[0].get_label("ner").value == "City" # but there is also one orgtype span and label labels: List[Label] = sentence.get_labels("orgtype") assert len(labels) == 1 assert labels[0].data_point.text == "Humboldt Universität zu Berlin" assert labels[0].value == "University" # and only one NER span spans: List[Span] = sentence.get_spans("orgtype") assert len(spans) == 1 assert spans[0].text == "Humboldt Universität zu Berlin" assert spans[0].get_label("orgtype").value == "University" # let's add the NER tag back sentence[0:4].add_label("ner", "Organization") # check if there are three labels with correct text and values labels: List[Label] = sentence.get_labels("ner") print(labels) assert len(labels) == 2 assert labels[0].data_point.text == "Humboldt Universität zu Berlin" assert labels[0].value == "Organization" assert labels[1].data_point.text == "Berlin" assert labels[1].value == "City" # check if there are two spans with correct text and values spans: List[Span] = sentence.get_spans("ner") assert len(spans) == 2 assert spans[0].text == "Humboldt Universität zu Berlin" assert spans[0].get_label("ner").value == "Organization" assert spans[0].get_label("orgtype").value == "University" assert len(spans[0].get_labels("ner")) == 1 assert spans[1].text == "Berlin" assert spans[1].get_label("ner").value == "City" # now remove all NER tags sentence.remove_labels("ner") assert len(sentence.get_labels("ner")) == 0 assert len(sentence.get_spans("ner")) == 0 assert len(sentence.get_spans("orgtype")) == 1 assert len(sentence.get_labels("orgtype")) == 1 assert len(sentence.labels) == 1 assert len(sentence[0:4].get_labels("ner")) == 0 assert len(sentence[0:4].get_labels("orgtype")) == 1 def test_relation_tags(): # set 3 labels for 2 spans (HU is tagged twice with different tags) sentence = Sentence("Humboldt Universität zu Berlin is located in Berlin .") # create two relation label Relation(sentence[0:4], sentence[7:8]).add_label("rel", "located in") Relation(sentence[0:2], sentence[3:4]).add_label("rel", "university of") Relation(sentence[0:2], sentence[3:4]).add_label("syntactic", "apposition") # there should be two relation labels labels: List[Label] = sentence.get_labels("rel") assert len(labels) == 2 assert labels[0].value == "located in" assert labels[1].value == "university of" # there should be one syntactic labels labels: List[Label] = sentence.get_labels("syntactic") assert len(labels) == 1 # there should be two relations, one with two and one with one label relations: List[Relation] = sentence.get_relations("rel") assert len(relations) == 2 assert len(relations[0].labels) == 1 assert len(relations[1].labels) == 2 def test_sentence_labels(): # example sentence sentence = Sentence("I love Berlin") sentence.add_label("sentiment", "positive") sentence.add_label("topic", "travelling") assert len(sentence.labels) == 2 assert len(sentence.get_labels("sentiment")) == 1 assert len(sentence.get_labels("topic")) == 1 # add another topic label sentence.add_label("topic", "travelling") assert len(sentence.labels) == 3 assert len(sentence.get_labels("sentiment")) == 1 assert len(sentence.get_labels("topic")) == 2 sentence.remove_labels("topic") assert len(sentence.labels) == 1 assert len(sentence.get_labels("sentiment")) == 1 assert len(sentence.get_labels("topic")) == 0 def test_mixed_labels(): # example sentence sentence = Sentence("I love New York") # has sentiment value sentence.add_label("sentiment", "positive") # has 4 part of speech tags sentence[1].add_label("pos", "verb") sentence[2].add_label("pos", "proper noun") sentence[3].add_label("pos", "proper noun") sentence[0].add_label("pos", "pronoun") # has 1 NER tag sentence[2:4].add_label("ner", "City") # should be in total 6 labels assert len(sentence.labels) == 6 assert len(sentence.get_labels("pos")) == 4 assert len(sentence.get_labels("sentiment")) == 1 assert len(sentence.get_labels("ner")) == 1 def test_data_point_equality(): # example sentence sentence = Sentence("George Washington went to Washington .") # add two NER labels sentence[0:2].add_label("span_ner", "PER") sentence[0:2].add_label("span_other", "Politician") sentence[4].add_label("ner", "LOC") sentence[4].add_label("other", "Village") # get the four labels ner_label = sentence.get_label("ner") other_label = sentence.get_label("other") span_ner_label = sentence.get_label("span_ner") span_other_label = sentence.get_label("span_other") # check that only two of the respective data points are equal assert ner_label.data_point == other_label.data_point assert span_ner_label.data_point == span_other_label.data_point assert ner_label.data_point != span_other_label.data_point assert other_label.data_point != span_ner_label.data_point
10,789
37.673835
80
py
flair
flair-master/tests/test_datasets.py
import copy import shutil import pytest import flair import flair.datasets from flair.data import MultiCorpus, Sentence from flair.datasets import ColumnCorpus from flair.datasets.sequence_labeling import ( ONTONOTES, JsonlCorpus, JsonlDataset, MultiFileJsonlCorpus, ) def test_load_imdb_data(tasks_base_path): # get training, test and dev data corpus = flair.datasets.ClassificationCorpus( tasks_base_path / "imdb", memory_mode="full", ) assert len(corpus.train) == 5 assert len(corpus.dev) == 5 assert len(corpus.test) == 5 def test_load_imdb_data_streaming(tasks_base_path): # get training, test and dev data corpus = flair.datasets.ClassificationCorpus( tasks_base_path / "imdb", memory_mode="disk", ) assert len(corpus.train) == 5 assert len(corpus.dev) == 5 assert len(corpus.test) == 5 def test_load_imdb_data_max_tokens(tasks_base_path): # get training, test and dev data corpus = flair.datasets.ClassificationCorpus(tasks_base_path / "imdb", memory_mode="full", truncate_to_max_tokens=3) assert len(corpus.train[0]) <= 3 assert len(corpus.dev[0]) <= 3 assert len(corpus.test[0]) <= 3 def test_load_imdb_data_streaming_max_tokens(tasks_base_path): # get training, test and dev data corpus = flair.datasets.ClassificationCorpus(tasks_base_path / "imdb", memory_mode="full", truncate_to_max_tokens=3) assert len(corpus.train[0]) <= 3 assert len(corpus.dev[0]) <= 3 assert len(corpus.test[0]) <= 3 def test_load_ag_news_data(tasks_base_path): # get training, test and dev data corpus = flair.datasets.ClassificationCorpus(tasks_base_path / "ag_news") assert len(corpus.train) == 10 assert len(corpus.dev) == 10 assert len(corpus.test) == 10 def test_load_sequence_labeling_data(tasks_base_path): # get training, test and dev data corpus = flair.datasets.ColumnCorpus(tasks_base_path / "fashion", column_format={0: "text", 2: "ner"}) assert len(corpus.train) == 6 assert len(corpus.dev) == 1 assert len(corpus.test) == 1 def test_load_sequence_labeling_whitespace_after(tasks_base_path): # get training, test and dev data corpus = flair.datasets.ColumnCorpus( tasks_base_path / "column_with_whitespaces", column_format={0: "text", 1: "ner", 2: "space-after"}, ) assert len(corpus.train) == 1 assert len(corpus.dev) == 1 assert len(corpus.test) == 1 assert corpus.train[0].to_tokenized_string() == "It is a German - owned firm ." assert corpus.train[0].to_plain_string() == "It is a German-owned firm." for token in corpus.train[0]: assert token.start_position is not None assert token.end_position is not None def test_load_column_corpus_options(tasks_base_path): # get training, test and dev data corpus = flair.datasets.ColumnCorpus( tasks_base_path / "column_corpus_options", column_format={0: "text", 1: "ner"}, column_delimiter="\t", skip_first_line=True, ) assert len(corpus.train) == 1 assert len(corpus.dev) == 1 assert len(corpus.test) == 1 assert corpus.train[0].to_tokenized_string() == "This is New Berlin" def test_load_span_data(tasks_base_path): # load column dataset with one entry dataset = flair.datasets.ColumnDataset( tasks_base_path / "span_labels" / "span_first.txt", column_name_map={0: "text", 1: "ner"}, ) assert len(dataset) == 1 assert dataset[0][2].text == "RAB" assert dataset[0][2].get_label("ner").value == "PARTA" # load column dataset with two entries dataset = flair.datasets.ColumnDataset( tasks_base_path / "span_labels" / "span_second.txt", column_name_map={0: "text", 1: "ner"}, ) assert len(dataset) == 2 assert dataset[1][2].text == "RAB" assert dataset[1][2].get_label("ner").value == "PARTA" # load column dataset with three entries dataset = flair.datasets.ColumnDataset( tasks_base_path / "span_labels" / "span_third.txt", column_name_map={0: "text", 1: "ner"}, ) assert len(dataset) == 3 assert dataset[2][2].text == "RAB" assert dataset[2][2].get_label("ner").value == "PARTA" def test_load_germeval_data(tasks_base_path): # get training, test and dev data corpus = flair.datasets.ColumnCorpus(tasks_base_path / "ner_german_germeval", column_format={0: "text", 2: "ner"}) assert len(corpus.train) == 2 assert len(corpus.dev) == 1 assert len(corpus.test) == 1 def test_load_ud_english_data(tasks_base_path): # get training, test and dev data corpus = flair.datasets.UD_ENGLISH(tasks_base_path) assert len(corpus.train) == 6 assert len(corpus.test) == 4 assert len(corpus.dev) == 2 # check if Token labels are correct sentence = corpus.train[0] assert sentence[0].text == "From" assert sentence[0].get_label("upos").value == "ADP" assert sentence[1].text == "the" assert sentence[1].get_label("upos").value == "DET" def test_load_up_english_data(tasks_base_path): # get training, test and dev data corpus = flair.datasets.UP_ENGLISH(tasks_base_path) assert len(corpus.train) == 4 assert len(corpus.test) == 2 assert len(corpus.dev) == 2 # check if Token labels for frames are correct sentence = corpus.dev[0] assert sentence[2].text == "AP" assert sentence[2].get_label("frame", zero_tag_value="no_label").value == "no_label" assert sentence[3].text == "comes" assert sentence[3].get_label("frame").value == "come.03" def test_load_no_dev_data(tasks_base_path): # get training, test and dev data corpus = flair.datasets.ColumnCorpus(tasks_base_path / "fashion_nodev", column_format={0: "text", 2: "ner"}) assert len(corpus.train) == 5 assert len(corpus.dev) == 1 assert len(corpus.test) == 1 def test_load_no_dev_data_explicit(tasks_base_path): # get training, test and dev data corpus = flair.datasets.ColumnCorpus( tasks_base_path / "fashion_nodev", column_format={0: "text", 2: "ner"}, train_file="train.tsv", test_file="test.tsv", ) assert len(corpus.train) == 5 assert len(corpus.dev) == 1 assert len(corpus.test) == 1 def test_multi_corpus(tasks_base_path): corpus_1 = flair.datasets.ColumnCorpus(tasks_base_path / "ner_german_germeval", column_format={0: "text", 2: "ner"}) corpus_2 = flair.datasets.ColumnCorpus(tasks_base_path / "fashion", column_format={0: "text", 2: "ner"}) # get two corpora as one corpus = MultiCorpus([corpus_1, corpus_2]) assert len(corpus.train) == 8 assert len(corpus.dev) == 2 assert len(corpus.test) == 2 def test_download_load_data(tasks_base_path): # get training, test and dev data for full English UD corpus from web corpus = flair.datasets.UD_ENGLISH() assert len(corpus.train) == 12544 assert len(corpus.dev) == 2001 assert len(corpus.test) == 2077 # clean up data directory shutil.rmtree(flair.cache_root / "datasets" / "ud_english") def _assert_conllu_dataset(dataset): sent1 = dataset[0] assert [label.data_point.text for label in sent1.get_labels("ner")] == ["Larry Page", "Sergey Brin", "Google"] assert [label.value for label in sent1.get_labels("ner")] == ["PER", "PER", "ORG"] assert [token.get_label("upos").value for token in sent1.tokens] == [ "PROPN", "PROPN", "CCONJ", "PROPN", "PROPN", "VERB", "PROPN", "PUNCT", ] assert [token.whitespace_after for token in sent1.tokens] == [ 1, 1, 1, 1, 1, 1, 0, 0, ] ner_spans1 = sent1.get_labels("ner") assert len(ner_spans1) == 3 upos_spans1 = sent1.get_labels("upos") assert len(upos_spans1) == 8 rels1 = sent1.get_labels("relation") assert len(rels1) == 2 assert [token.idx for token in rels1[1].data_point.first] == [7] assert [token.idx for token in rels1[1].data_point.second] == [4, 5] sent3 = dataset[2] ner_labels3 = sent3.get_labels("ner") assert len(ner_labels3) == 3 upos_labels3 = sent3.get_labels("upos") assert len(upos_labels3) == 11 rels3 = sent3.get_labels("relation") assert len(rels3) == 1 assert [token.idx for token in rels3[0].data_point.first] == [6] assert [token.idx for token in rels3[0].data_point.second] == [1, 2] def test_load_conllu_corpus(tasks_base_path): corpus = ColumnCorpus( tasks_base_path / "conllu", train_file="train.conllu", dev_file="train.conllu", test_file="train.conllu", in_memory=False, column_format={1: "text", 2: "upos", 3: "ner", 4: "feats"}, ) assert len(corpus.train) == 4 assert len(corpus.dev) == 4 assert len(corpus.test) == 4 _assert_conllu_dataset(corpus.train) def test_load_conllu_corpus_in_memory(tasks_base_path): corpus = ColumnCorpus( tasks_base_path / "conllu", train_file="train.conllu", dev_file="train.conllu", test_file="train.conllu", column_format={1: "text", 2: "upos", 3: "ner", 4: "feats"}, in_memory=True, ) assert len(corpus.train) == 4 assert len(corpus.dev) == 4 assert len(corpus.test) == 4 _assert_conllu_dataset(corpus.train) def test_load_conllu_plus_corpus(tasks_base_path): corpus = ColumnCorpus( tasks_base_path / "conllu", train_file="train.conllup", dev_file="train.conllup", test_file="train.conllup", column_format={1: "text", 2: "upos", 3: "ner", 4: "feats"}, in_memory=False, ) assert len(corpus.train) == 4 assert len(corpus.dev) == 4 assert len(corpus.test) == 4 _assert_conllu_dataset(corpus.train) def test_load_conllu_corpus_plus_in_memory(tasks_base_path): corpus = ColumnCorpus( tasks_base_path / "conllu", train_file="train.conllup", dev_file="train.conllup", test_file="train.conllup", column_format={1: "text", 2: "upos", 3: "ner", 4: "feats"}, in_memory=True, ) assert len(corpus.train) == 4 assert len(corpus.dev) == 4 assert len(corpus.test) == 4 _assert_conllu_dataset(corpus.train) def _assert_universal_dependencies_conllu_dataset(dataset): sent1: Sentence = dataset[0] assert [token.whitespace_after for token in sent1.tokens] == [ 1, 1, 1, 1, 0, 0, ] assert len(sent1.get_labels("Number")) == 4 assert sent1[1].get_labels("Number")[0].value == "Plur" assert sent1[1].get_labels("Person")[0].value == "3" assert sent1[1].get_labels("Tense")[0].value == "Pres" # assert [token.get_tag("head").value for token in sent1.tokens] == [ # "2", # "0", # "4", # "2", # "2", # "2", # ] assert [token.get_label("deprel").value for token in sent1.tokens] == [ "nsubj", "root", "cc", "conj", "obj", "punct", ] def test_load_universal_dependencies_conllu_corpus(tasks_base_path): # This test only covers basic universal dependencies datasets. # For example, multi-word tokens or the "deps" column sentence annotations are not supported yet. # Here, we use the default token annotation fields. corpus = ColumnCorpus( tasks_base_path / "conllu", train_file="universal_dependencies.conllu", dev_file="universal_dependencies.conllu", test_file="universal_dependencies.conllu", column_format={ 1: "text", 2: "lemma", 3: "upos", 4: "pos", 5: "feats", 6: "head", 7: "deprel", 8: "deps", 9: "misc", }, ) assert len(corpus.train) == 1 assert len(corpus.dev) == 1 assert len(corpus.test) == 1 _assert_universal_dependencies_conllu_dataset(corpus.train) def test_hipe_2022_corpus(tasks_base_path): # This test covers the complete HIPE 2022 dataset. # https://github.com/hipe-eval/HIPE-2022-data # Includes variant with document separator, and all versions of the dataset. # We have manually checked, that these numbers are correct: hipe_stats = { "v1.0": { "ajmc": { "de": {"sample": {"sents": 119, "docs": 8, "labels": ["date", "loc", "pers", "scope", "work"]}}, "en": {"sample": {"sents": 83, "docs": 5, "labels": ["date", "loc", "pers", "scope", "work"]}}, }, "hipe2020": { "de": { "train": { "sents": 3470 + 2, # 2 sentences with missing EOS marker "docs": 103, "labels": ["loc", "org", "pers", "prod", "time"], }, "dev": {"sents": 1202, "docs": 33, "labels": ["loc", "org", "pers", "prod", "time"]}, }, "en": {"dev": {"sents": 1045, "docs": 80, "labels": ["loc", "org", "pers", "prod", "time"]}}, "fr": { "train": {"sents": 5743, "docs": 158, "labels": ["loc", "org", "pers", "prod", "time", "comp"]}, "dev": {"sents": 1244, "docs": 43, "labels": ["loc", "org", "pers", "prod", "time"]}, }, }, "letemps": { "fr": { "train": {"sents": 14051, "docs": 414, "labels": ["loc", "org", "pers"]}, "dev": {"sents": 1341, "docs": 51, "labels": ["loc", "org", "pers"]}, } }, "newseye": { # +1 offset, because of missing EOS marker at EOD "de": { "train": {"sents": 23646 + 1, "docs": 11, "labels": ["HumanProd", "LOC", "ORG", "PER"]}, "dev": {"sents": 1110 + 1, "docs": 12, "labels": ["HumanProd", "LOC", "ORG", "PER"]}, "dev2": {"sents": 1541 + 1, "docs": 12, "labels": ["HumanProd", "LOC", "ORG", "PER"]}, }, "fi": { "train": {"sents": 1141 + 1, "docs": 24, "labels": ["HumanProd", "LOC", "ORG", "PER"]}, "dev": {"sents": 140 + 1, "docs": 24, "labels": ["HumanProd", "LOC", "ORG", "PER"]}, "dev2": {"sents": 104 + 1, "docs": 21, "labels": ["HumanProd", "LOC", "ORG", "PER"]}, }, "fr": { "train": {"sents": 7106 + 1, "docs": 35, "labels": ["HumanProd", "LOC", "ORG", "PER"]}, "dev": {"sents": 662 + 1, "docs": 35, "labels": ["HumanProd", "LOC", "ORG", "PER"]}, "dev2": {"sents": 1016 + 1, "docs": 35, "labels": ["HumanProd", "LOC", "ORG", "PER"]}, }, "sv": { "train": {"sents": 1063 + 1, "docs": 21, "labels": ["HumanProd", "LOC", "ORG", "PER"]}, "dev": {"sents": 126 + 1, "docs": 21, "labels": ["HumanProd", "LOC", "ORG", "PER"]}, "dev2": {"sents": 136 + 1, "docs": 21, "labels": ["HumanProd", "LOC", "ORG", "PER"]}, }, }, "sonar": { "de": { "dev": { "sents": 1603 + 10, # 10 sentences with missing EOS marker "docs": 10, "labels": ["LOC", "ORG", "PER"], } } }, "topres19th": { "en": { "train": {"sents": 5874, "docs": 309, "labels": ["BUILDING", "LOC", "STREET"]}, "dev": {"sents": 646, "docs": 34, "labels": ["BUILDING", "LOC", "STREET"]}, } }, } } hipe_stats["v2.0"] = copy.deepcopy(hipe_stats["v1.0"]) hipe_stats["v2.0"]["ajmc"] = { "de": { "train": { "sents": 1022 + 2, # 2 sentences with missing EOS marker "docs": 76, "labels": ["date", "loc", "object", "pers", "scope", "work"], }, "dev": {"sents": 192, "docs": 14, "labels": ["loc", "object", "pers", "scope", "work"]}, }, "en": { "train": { "sents": 1153 + 1, # 1 sentence with missing EOS marker "docs": 60, "labels": ["date", "loc", "object", "pers", "scope", "work"], }, "dev": { "sents": 251 + 1, # 1 sentence with missing EOS marker "docs": 14, "labels": ["date", "loc", "pers", "scope", "work"], }, }, "fr": { "train": { "sents": 893 + 1, # 1 sentence with missing EOS marker "docs": 72, "labels": ["date", "loc", "object", "pers", "scope", "work"], }, "dev": { "sents": 201 + 1, # 1 sentence with missing EOS marker "docs": 17, "labels": ["pers", "scope", "work"], }, }, } hipe_stats["v2.0"]["newseye"]["de"] = { "train": {"sents": 20839 + 1, "docs": 7, "labels": ["HumanProd", "LOC", "ORG", "PER"]} # missing EOD marker } hipe_stats["v2.0"]["sonar"] = { "de": { "dev": { "sents": 816 + 10, # 9 sentences with missing EOS marker + missing EOD "docs": 10, "labels": ["LOC", "ORG", "PER"], } } } hipe_stats["v2.1"] = copy.deepcopy(hipe_stats["v2.0"]) hipe_stats["v2.1"]["hipe2020"]["fr"]["train"] = { "sents": 5743, "docs": 158, "labels": ["loc", "org", "pers", "prod", "time"], } # Test data for v2.1 release hipe_stats["v2.1"]["ajmc"]["de"]["test"] = { "sents": 224, "docs": 16, "labels": ["loc", "object", "pers", "scope", "work"], } hipe_stats["v2.1"]["ajmc"]["en"]["test"] = { "sents": 238, "docs": 13, "labels": ["date", "loc", "pers", "scope", "work"], } hipe_stats["v2.1"]["ajmc"]["fr"]["test"] = { "sents": 188 + 1, # 1 sentence with missing EOS marker "docs": 15, "labels": ["date", "loc", "pers", "scope", "work"], } hipe_stats["v2.1"]["hipe2020"]["de"]["test"] = { "sents": 1215 + 2, # 2 sentences with missing EOS marker "docs": 49, "labels": ["loc", "org", "pers", "prod", "time"], } hipe_stats["v2.1"]["hipe2020"]["en"]["test"] = { "sents": 553, "docs": 46, "labels": ["loc", "org", "pers", "prod", "time"], } hipe_stats["v2.1"]["hipe2020"]["fr"]["test"] = { "sents": 1462, "docs": 43, "labels": ["loc", "org", "pers", "prod", "time"], } hipe_stats["v2.1"]["letemps"]["fr"]["test"] = {"sents": 2381, "docs": 51, "labels": ["loc", "org", "pers"]} hipe_stats["v2.1"]["newseye"]["de"]["test"] = { "sents": 3336 + 1, # 1 missing EOD marker "docs": 13, "labels": ["HumanProd", "LOC", "ORG", "PER"], } hipe_stats["v2.1"]["newseye"]["fi"]["test"] = { "sents": 390 + 1, # 1 missing EOD marker "docs": 24, "labels": ["HumanProd", "LOC", "ORG", "PER"], } hipe_stats["v2.1"]["newseye"]["fr"]["test"] = { "sents": 2534 + 1, # 1 missing EOD marker "docs": 35, "labels": ["HumanProd", "LOC", "ORG", "PER"], } hipe_stats["v2.1"]["newseye"]["sv"]["test"] = { "sents": 342 + 1, # 1 missing EOD marker "docs": 21, "labels": ["HumanProd", "LOC", "ORG", "PER"], } hipe_stats["v2.1"]["sonar"]["de"]["test"] = { "sents": 807 + 8 + 1, # 8 missing EOS marker + missing EOD "docs": 10, "labels": ["LOC", "ORG", "PER"], } hipe_stats["v2.1"]["topres19th"]["en"]["test"] = { "sents": 2001, "docs": 112, "labels": ["BUILDING", "LOC", "STREET"], } def test_hipe_2022(dataset_version="v2.1", add_document_separator=True): for dataset_name, languages in hipe_stats[dataset_version].items(): for language in languages: splits = languages[language] corpus = flair.datasets.NER_HIPE_2022( version=dataset_version, dataset_name=dataset_name, language=language, dev_split_name="dev", add_document_separator=add_document_separator, ) for split_name, stats in splits.items(): split_description = f"{dataset_name}@{dataset_version}/{language}#{split_name}" current_sents = stats["sents"] current_docs = stats["docs"] current_labels = set(stats["labels"]) total_sentences = current_sents + current_docs if add_document_separator else stats["sents"] if split_name == "train": assert ( len(corpus.train) == total_sentences ), f"Sentence count mismatch for {split_description}: {len(corpus.train)} vs. {total_sentences}" gold_labels = set(corpus.make_label_dictionary(label_type="ner").get_items()) assert ( current_labels == gold_labels ), f"Label mismatch for {split_description}: {current_labels} vs. {gold_labels}" elif split_name in ["dev", "sample"]: assert ( len(corpus.dev) == total_sentences ), f"Sentence count mismatch for {split_description}: {len(corpus.dev)} vs. {total_sentences}" corpus._train = corpus._dev gold_labels = set(corpus.make_label_dictionary(label_type="ner").get_items()) assert ( current_labels == gold_labels ), f"Label mismatch for {split_description}: {current_labels} vs. {gold_labels}" elif split_name == "dev2": corpus = flair.datasets.NER_HIPE_2022( version=dataset_version, dataset_name=dataset_name, language=language, dev_split_name="dev2", add_document_separator=add_document_separator, ) corpus._train = corpus._dev gold_labels = set(corpus.make_label_dictionary(label_type="ner").get_items()) assert ( len(corpus.dev) == total_sentences ), f"Sentence count mismatch for {split_description}: {len(corpus.dev)} vs. {total_sentences}" assert ( current_labels == gold_labels ), f"Label mismatch for {split_description}: {current_labels} vs. {gold_labels}" test_hipe_2022(dataset_version="v1.0", add_document_separator=True) test_hipe_2022(dataset_version="v1.0", add_document_separator=False) test_hipe_2022(dataset_version="v2.0", add_document_separator=True) test_hipe_2022(dataset_version="v2.0", add_document_separator=False) test_hipe_2022(dataset_version="v2.1", add_document_separator=True) test_hipe_2022(dataset_version="v2.1", add_document_separator=False) def test_icdar_europeana_corpus(tasks_base_path): # This test covers the complete ICDAR Europeana corpus: # https://github.com/stefan-it/historic-domain-adaptation-icdar gold_stats = {"fr": {"train": 7936, "dev": 992, "test": 992}, "nl": {"train": 5777, "dev": 722, "test": 723}} def check_number_sentences(reference: int, actual: int, split_name: str): assert actual == reference, f"Mismatch in number of sentences for {split_name} split" for language in ["fr", "nl"]: corpus = flair.datasets.NER_ICDAR_EUROPEANA(language=language) check_number_sentences(len(corpus.train), gold_stats[language]["train"], "train") check_number_sentences(len(corpus.dev), gold_stats[language]["dev"], "dev") check_number_sentences(len(corpus.test), gold_stats[language]["test"], "test") def test_masakhane_corpus(tasks_base_path): # This test covers the complete MasakhaNER dataset, including support for v1 and v2. supported_versions = ["v1", "v2"] supported_languages = { "v1": ["amh", "hau", "ibo", "kin", "lug", "luo", "pcm", "swa", "wol", "yor"], "v2": [ "bam", "bbj", "ewe", "fon", "hau", "ibo", "kin", "lug", "mos", "pcm", "nya", "sna", "swa", "tsn", "twi", "wol", "xho", "yor", "zul", ], } masakhane_stats = { "v1": { "amh": {"train": 1750, "dev": 250, "test": 500}, "hau": {"train": 1912, "dev": 276, "test": 552}, "ibo": {"train": 2235, "dev": 320, "test": 638}, "kin": { "train": 2116, "dev": 302, "test": 605, }, "lug": {"train": 1428, "dev": 200, "test": 407}, "luo": {"train": 644, "dev": 92, "test": 186}, "pcm": {"train": 2124, "dev": 306, "test": 600}, "swa": {"train": 2109, "dev": 300, "test": 604}, "wol": {"train": 1871, "dev": 267, "test": 539}, "yor": {"train": 2171, "dev": 305, "test": 645}, }, "v2": { "bam": {"train": 4462, "dev": 638, "test": 1274}, "bbj": {"train": 3384, "dev": 483, "test": 966}, "ewe": {"train": 3505, "dev": 501, "test": 1001}, "fon": {"train": 4343, "dev": 621, "test": 1240}, "hau": {"train": 5716, "dev": 816, "test": 1633}, "ibo": {"train": 7634, "dev": 1090, "test": 2181}, "kin": {"train": 7825, "dev": 1118, "test": 2235}, "lug": {"train": 4942, "dev": 706, "test": 1412}, "mos": {"train": 4532, "dev": 648, "test": 1294}, "pcm": {"train": 5646, "dev": 806, "test": 1613}, "nya": {"train": 6250, "dev": 893, "test": 1785}, "sna": {"train": 6207, "dev": 887, "test": 1773}, "swa": {"train": 6593, "dev": 942, "test": 1883}, "tsn": {"train": 3489, "dev": 499, "test": 996}, "twi": {"train": 4240, "dev": 605, "test": 1211}, "wol": {"train": 4593, "dev": 656, "test": 1312}, "xho": {"train": 5718, "dev": 817, "test": 1633}, "yor": {"train": 6876, "dev": 983, "test": 1964}, "zul": {"train": 5848, "dev": 836, "test": 1670}, }, } def check_number_sentences(reference: int, actual: int, split_name: str, language: str, version: str): assert actual == reference, f"Mismatch in number of sentences for {language}@{version}/{split_name}" for version in supported_versions: for language in supported_languages[version]: corpus = flair.datasets.NER_MASAKHANE(languages=language, version=version) gold_stats = masakhane_stats[version][language] check_number_sentences(len(corpus.train), gold_stats["train"], "train", language, version) check_number_sentences(len(corpus.dev), gold_stats["dev"], "dev", language, version) check_number_sentences(len(corpus.test), gold_stats["test"], "test", language, version) def test_nermud_corpus(tasks_base_path): # This test covers the NERMuD dataset. Official stats can be found here: # https://github.com/dhfbk/KIND/tree/main/evalita-2023 gold_stats = { "WN": {"train": 10912, "dev": 2594}, "FIC": {"train": 11423, "dev": 1051}, "ADG": {"train": 5147, "dev": 1122}, } def check_number_sentences(reference: int, actual: int, split_name: str): assert actual == reference, f"Mismatch in number of sentences for {split_name} split" for domain, stats in gold_stats.items(): corpus = flair.datasets.NER_NERMUD(domains=domain) check_number_sentences(len(corpus.train), stats["train"], "train") check_number_sentences(len(corpus.dev), stats["dev"], "dev") def test_multi_file_jsonl_corpus_should_use_label_type(tasks_base_path): corpus = MultiFileJsonlCorpus( train_files=[tasks_base_path / "jsonl/train.jsonl"], dev_files=[tasks_base_path / "jsonl/testa.jsonl"], test_files=[tasks_base_path / "jsonl/testb.jsonl"], label_type="pos", ) for sentence in corpus.get_all_sentences(): assert sentence.has_label("pos") assert not sentence.has_label("ner") def test_jsonl_corpus_should_use_label_type(tasks_base_path): corpus = JsonlCorpus(tasks_base_path / "jsonl", label_type="pos") for sentence in corpus.get_all_sentences(): assert sentence.has_label("pos") assert not sentence.has_label("ner") def test_jsonl_dataset_should_use_label_type(tasks_base_path): """Tests whether the dataset respects the label_type parameter.""" dataset = JsonlDataset(tasks_base_path / "jsonl" / "train.jsonl", label_type="pos") # use other type for sentence in dataset.sentences: assert sentence.has_label("pos") assert not sentence.has_label("ner") def test_reading_jsonl_dataset_should_be_successful(tasks_base_path): """Tests reading a JsonlDataset containing multiple tagged entries.""" dataset = JsonlDataset(tasks_base_path / "jsonl" / "train.jsonl") assert len(dataset.sentences) == 5 assert len(dataset.sentences[0].get_labels("ner")) == 1 assert dataset.sentences[0][2:4].get_label("ner").value == "LOC" def test_simple_folder_jsonl_corpus_should_load(tasks_base_path): corpus = JsonlCorpus(tasks_base_path / "jsonl") assert len(corpus.get_all_sentences()) == 11 def test_jsonl_corpus_loads_spans(tasks_base_path): corpus = JsonlCorpus(tasks_base_path / "jsonl") assert corpus.train is not None example = corpus.train[0] assert len(example.get_spans("ner")) > 0 def test_ontonotes_download(): from urllib.parse import urlparse res = urlparse(ONTONOTES.archive_url) assert all([res.scheme, res.netloc]) def test_ontonotes_extraction(tasks_base_path): import os import tempfile from flair.file_utils import unpack_file ontonotes_path = tasks_base_path / "ontonotes" with tempfile.TemporaryDirectory() as tmp_dir: unpack_file(ontonotes_path / "tiny-conll-2012.zip", tmp_dir, "zip", True) assert "conll-2012" in os.listdir(tmp_dir) corpus = ONTONOTES(base_path=tmp_dir) label_dictionary = corpus.make_label_dictionary("ner") assert len(label_dictionary) == 14 assert label_dictionary.span_labels domain_specific_corpus = ONTONOTES(base_path=tmp_dir, domain=["bc"]) assert len(corpus.train) > len(domain_specific_corpus.train) TRAIN_FILE = "tests/resources/tasks/jsonl/train.jsonl" TESTA_FILE = "tests/resources/tasks/jsonl/testa.jsonl" TESTB_FILE = "tests/resources/tasks/jsonl/testa.jsonl" @pytest.mark.parametrize( ("train_files", "dev_files", "test_files", "expected_size"), [ ( [TRAIN_FILE], [TESTA_FILE], [TESTB_FILE], 11, ), ( [TRAIN_FILE], [], [TESTB_FILE], 8, ), ( [TRAIN_FILE], [], None, 5, ), ( None, [TESTA_FILE], None, 3, ), ( [TRAIN_FILE, TESTA_FILE], [TESTA_FILE], [TESTB_FILE], 14, ), ], ) def test_corpus_with_single_files_should_load(train_files, dev_files, test_files, expected_size): corpus = MultiFileJsonlCorpus(train_files, dev_files, test_files) assert len(corpus.get_all_sentences()) == expected_size def test_empty_corpus_should_raise_error(): with pytest.raises(RuntimeError) as err: MultiFileJsonlCorpus(None, None, None) assert str(err.value) == "No data provided when initializing corpus object."
32,776
34.168455
120
py
flair
flair-master/tests/test_tokenize_sentence.py
from typing import List import pytest import flair from flair.data import Sentence, Token from flair.splitter import ( NewlineSentenceSplitter, NoSentenceSplitter, SciSpacySentenceSplitter, SegtokSentenceSplitter, SpacySentenceSplitter, TagSentenceSplitter, ) from flair.tokenization import ( JapaneseTokenizer, SciSpacyTokenizer, SegtokTokenizer, SpaceTokenizer, SpacyTokenizer, TokenizerWrapper, ) def test_create_sentence_on_empty_string(): sentence: Sentence = Sentence("") assert len(sentence.tokens) == 0 def test_create_sentence_with_newline(): sentence: Sentence = Sentence(["I", "\t", "ich", "\n", "you", "\t", "du", "\n"]) assert len(sentence.tokens) == 8 assert sentence.tokens[3].text == "\n" sentence: Sentence = Sentence("I \t ich \n you \t du \n", use_tokenizer=False) assert len(sentence.tokens) == 8 assert sentence.tokens[0].start_position == 0 assert sentence.tokens[3].text == "\n" def test_create_sentence_with_extra_whitespace(): sentence: Sentence = Sentence("I love Berlin .") assert len(sentence.tokens) == 4 assert sentence.get_token(1).text == "I" assert sentence.get_token(2).text == "love" assert sentence.get_token(3).text == "Berlin" assert sentence.get_token(4).text == "." def test_create_sentence_difficult_encoding(): text = "so out of the norm ❤ ️ enjoyed every moment️" sentence = Sentence(text) assert len(sentence) == 9 text = ( "equivalently , accumulating the logs as :( 6 ) sl = 1N ∑ t = 1Nlogp " "( Ll | xt \u200b , θ ) where " "p ( Ll | xt \u200b , θ ) represents the class probability output" ) sentence = Sentence(text) assert len(sentence) == 37 text = "This guy needs his own show on Discivery Channel ! " sentence = Sentence(text) assert len(sentence) == 10 text = "n't have new vintages." sentence = Sentence(text, use_tokenizer=True) assert len(sentence) == 5 def test_create_sentence_word_by_word(): token1: Token = Token("Munich") token2: Token = Token("and") token3: Token = Token("Berlin") token4: Token = Token("are") token5: Token = Token("nice") sentence: Sentence = Sentence([token1, token2, token3, token4, token5, Token("cities"), Token(".")]) assert sentence.to_tokenized_string() == "Munich and Berlin are nice cities ." def test_create_sentence_pretokenized(): pretoks = ["The", "grass", "is", "green", "."] sent = Sentence(pretoks) for i, token in enumerate(sent): assert token.text == pretoks[i] def test_create_sentence_without_tokenizer(): sentence: Sentence = Sentence("I love Berlin.", use_tokenizer=False) assert len(sentence.tokens) == 3 assert sentence.tokens[0].start_position == 0 assert sentence.tokens[0].text == "I" assert sentence.tokens[1].start_position == 2 assert sentence.tokens[1].text == "love" assert sentence.tokens[2].start_position == 7 assert sentence.tokens[2].text == "Berlin." def test_create_sentence_with_default_tokenizer(): sentence: Sentence = Sentence("I love Berlin.", use_tokenizer=True) assert len(sentence.tokens) == 4 assert sentence.tokens[0].start_position == 0 assert sentence.tokens[0].text == "I" assert sentence.tokens[1].start_position == 2 assert sentence.tokens[1].text == "love" assert sentence.tokens[2].start_position == 7 assert sentence.tokens[2].text == "Berlin" assert sentence.tokens[3].start_position == 13 assert sentence.tokens[3].text == "." def test_create_sentence_with_segtok(): sentence: Sentence = Sentence("I love Berlin.", use_tokenizer=SegtokTokenizer()) assert len(sentence.tokens) == 4 assert sentence.tokens[0].text == "I" assert sentence.tokens[1].text == "love" assert sentence.tokens[2].text == "Berlin" assert sentence.tokens[3].text == "." def test_create_sentence_with_custom_tokenizer(): sentence: Sentence = Sentence("I love Berlin.", use_tokenizer=TokenizerWrapper(no_op_tokenizer)) assert len(sentence.tokens) == 1 assert sentence.tokens[0].start_position == 0 assert sentence.tokens[0].text == "I love Berlin." @pytest.mark.skip(reason="SpacyTokenizer needs optional requirements, so we skip the test by default") def test_create_sentence_with_spacy_tokenizer(): sentence: Sentence = Sentence("I love Berlin.", use_tokenizer=SpacyTokenizer("en_core_sci_sm")) assert len(sentence.tokens) == 4 assert sentence.tokens[0].start_position == 0 assert sentence.tokens[0].text == "I" assert sentence.tokens[1].start_position == 2 assert sentence.tokens[1].text == "love" assert sentence.tokens[2].start_position == 7 assert sentence.tokens[2].text == "Berlin" assert sentence.tokens[3].start_position == 13 assert sentence.tokens[3].text == "." def test_create_sentence_using_japanese_tokenizer(): sentence: Sentence = Sentence("私はベルリンが好き", use_tokenizer=JapaneseTokenizer("janome")) assert len(sentence.tokens) == 5 assert sentence.tokens[0].text == "私" assert sentence.tokens[1].text == "は" assert sentence.tokens[2].text == "ベルリン" assert sentence.tokens[3].text == "が" assert sentence.tokens[4].text == "好き" @pytest.mark.skip(reason="SciSpacyTokenizer need optional requirements, so we skip the test by default") def test_create_sentence_using_scispacy_tokenizer(): sentence: Sentence = Sentence( "Spinal and bulbar muscular atrophy (SBMA) is an inherited motor neuron", use_tokenizer=SciSpacyTokenizer(), ) assert len(sentence.tokens) == 13 assert sentence.tokens[0].text == "Spinal" assert sentence.tokens[1].text == "and" assert sentence.tokens[2].text == "bulbar" assert sentence.tokens[3].text == "muscular" assert sentence.tokens[4].text == "atrophy" assert sentence.tokens[5].text == "(" assert sentence.tokens[6].text == "SBMA" assert sentence.tokens[7].text == ")" assert sentence.tokens[8].text == "is" assert sentence.tokens[9].text == "an" assert sentence.tokens[10].text == "inherited" assert sentence.tokens[11].text == "motor" assert sentence.tokens[12].text == "neuron" assert sentence.tokens[0].start_position == 0 assert sentence.tokens[1].start_position == 7 assert sentence.tokens[2].start_position == 11 assert sentence.tokens[3].start_position == 18 assert sentence.tokens[4].start_position == 27 assert sentence.tokens[5].start_position == 35 assert sentence.tokens[6].start_position == 36 assert sentence.tokens[7].start_position == 40 assert sentence.tokens[8].start_position == 42 assert sentence.tokens[9].start_position == 45 assert sentence.tokens[10].start_position == 48 assert sentence.tokens[11].start_position == 58 assert sentence.tokens[12].start_position == 64 assert sentence.tokens[4].whitespace_after == 1 assert sentence.tokens[5].whitespace_after != 1 assert sentence.tokens[6].whitespace_after != 1 assert sentence.tokens[7].whitespace_after == 1 def test_split_text_segtok(): segtok_splitter = SegtokSentenceSplitter() sentences = segtok_splitter.split("I love Berlin. Berlin is a great city.") assert len(sentences) == 2 assert sentences[0].start_position == 0 assert len(sentences[0].tokens) == 4 assert sentences[1].start_position == 15 assert len(sentences[1].tokens) == 6 segtok_splitter = SegtokSentenceSplitter(tokenizer=TokenizerWrapper(no_op_tokenizer)) sentences = segtok_splitter.split("I love Berlin. Berlin is a great city.") assert len(sentences) == 2 assert sentences[0].start_position == 0 assert len(sentences[0].tokens) == 1 assert sentences[1].start_position == 15 assert len(sentences[1].tokens) == 1 def test_split_text_nosplit(): no_splitter = NoSentenceSplitter() sentences = no_splitter.split("I love Berlin") assert len(sentences) == 1 assert sentences[0].start_position == 0 assert len(sentences[0].tokens) == 3 no_splitter = NoSentenceSplitter(TokenizerWrapper(no_op_tokenizer)) sentences = no_splitter.split("I love Berlin") assert len(sentences) == 1 assert sentences[0].start_position == 0 assert len(sentences[0].tokens) == 1 def test_split_text_on_tag(): tag_splitter = TagSentenceSplitter(tag="#!") sentences = tag_splitter.split("I love Berlin#!Me too") assert len(sentences) == 2 assert sentences[0].start_position == 0 assert len(sentences[0].tokens) == 3 assert sentences[1].start_position == 15 assert len(sentences[1].tokens) == 2 tag_splitter = TagSentenceSplitter(tag="#!", tokenizer=TokenizerWrapper(no_op_tokenizer)) sentences = tag_splitter.split("I love Berlin#!Me too") assert len(sentences) == 2 assert sentences[0].start_position == 0 assert len(sentences[0].tokens) == 1 assert sentences[1].start_position == 15 assert len(sentences[1].tokens) == 1 sentences = tag_splitter.split("I love Berlin Me too") assert len(sentences) == 1 sentences = tag_splitter.split("I love Berlin#!#!Me too") assert len(sentences) == 2 sentences = tag_splitter.split("I love Berl#! #!inMe too") assert len(sentences) == 2 def test_split_text_on_newline(): newline_splitter = NewlineSentenceSplitter() sentences = newline_splitter.split("I love Berlin\nMe too") assert len(sentences) == 2 assert sentences[0].start_position == 0 assert len(sentences[0].tokens) == 3 assert sentences[0].start_position == 0 assert len(sentences[1].tokens) == 2 newline_splitter = NewlineSentenceSplitter(tokenizer=TokenizerWrapper(no_op_tokenizer)) sentences = newline_splitter.split("I love Berlin\nMe too") assert len(sentences) == 2 assert len(sentences[0].tokens) == 1 assert sentences[1].start_position == 14 assert len(sentences[1].tokens) == 1 sentences = newline_splitter.split("I love Berlin Me too") assert len(sentences) == 1 sentences = newline_splitter.split("I love Berlin\n\nMe too") assert len(sentences) == 2 sentences = newline_splitter.split("I love Berlin\n \nMe too") assert len(sentences) == 2 @pytest.mark.skip(reason="SpacySentenceSplitter need optional requirements, so we skip the test by default") def test_split_text_spacy(): spacy_splitter = SpacySentenceSplitter("en_core_sci_sm") sentences = spacy_splitter.split("This a sentence. And here is another one.") assert len(sentences) == 2 assert sentences[0].start_position == 0 assert len(sentences[0].tokens) == 4 assert sentences[1].start_position == 17 assert len(sentences[1].tokens) == 6 sentences = spacy_splitter.split("VF inhibits something. ACE-dependent (GH+) issuses too.") assert len(sentences) == 2 assert sentences[0].start_position == 0 assert len(sentences[0].tokens) == 4 assert sentences[1].start_position == 23 assert len(sentences[1].tokens) == 7 spacy_splitter = SpacySentenceSplitter("en_core_sci_sm", tokenizer=TokenizerWrapper(no_op_tokenizer)) sentences = spacy_splitter.split("This a sentence. And here is another one.") assert len(sentences) == 2 assert sentences[0].start_position == 0 assert len(sentences[0].tokens) == 1 assert sentences[1].start_position == 17 assert len(sentences[1].tokens) == 1 @pytest.mark.skip(reason="SciSpacySentenceSplitter need optional requirements, so we skip the test by default") def test_split_text_scispacy(): scispacy_splitter = SciSpacySentenceSplitter() sentences = scispacy_splitter.split("VF inhibits something. ACE-dependent (GH+) issuses too.") assert len(sentences) == 2 assert sentences[0].start_position == 0 assert len(sentences[0].tokens) == 4 assert sentences[1].start_position == 23 assert len(sentences[1].tokens) == 9 def test_print_sentence_tokenized(): sentence: Sentence = Sentence("I love Berlin.", use_tokenizer=SegtokTokenizer()) assert sentence.to_tokenized_string() == "I love Berlin ." def test_print_original_text(): text = ": nation on" sentence = Sentence(text) assert text == sentence.to_original_text() text = ": nation on" sentence = Sentence(text, use_tokenizer=SegtokTokenizer()) assert text == sentence.to_original_text() text = "I love Berlin." sentence = Sentence(text) assert text == sentence.to_original_text() text = ( 'Schartau sagte dem " Tagesspiegel " vom Freitag , Fischer sei " ' "in einer Weise aufgetreten , die alles andere als überzeugend " 'war " .' ) sentence = Sentence(text) assert text == sentence.to_original_text() text = ( 'Schartau sagte dem " Tagesspiegel " vom Freitag , Fischer sei " ' "in einer Weise aufgetreten , die alles andere als überzeugend " 'war " .' ) sentence = Sentence(text, use_tokenizer=SegtokTokenizer()) assert text == sentence.to_original_text() def test_print_sentence_plain(tasks_base_path): sentence: Sentence = Sentence("I love Berlin.", use_tokenizer=SegtokTokenizer()) assert sentence.to_plain_string() == "I love Berlin." corpus = flair.datasets.NER_GERMAN_GERMEVAL(base_path=tasks_base_path) sentence = corpus.train[0] sentence.infer_space_after() assert ( sentence.to_tokenized_string() == 'Schartau sagte dem " Tagesspiegel " vom Freitag , Fischer sei " in ' "einer Weise aufgetreten , " 'die alles andere als überzeugend war " .' ) assert ( sentence.to_plain_string() == 'Schartau sagte dem "Tagesspiegel" vom Freitag, Fischer sei "in einer ' "Weise aufgetreten, die " 'alles andere als überzeugend war".' ) sentence = corpus.train[1] sentence.infer_space_after() assert ( sentence.to_tokenized_string() == "Firmengründer Wolf Peter Bree arbeitete Anfang der siebziger Jahre als " "Möbelvertreter , als er einen fliegenden Händler aus dem Libanon traf ." ) assert ( sentence.to_plain_string() == "Firmengründer Wolf Peter Bree arbeitete Anfang der siebziger Jahre als " "Möbelvertreter, als er einen fliegenden Händler aus dem Libanon traf." ) def test_infer_space_after(): sentence: Sentence = Sentence([Token("xyz"), Token('"'), Token("abc"), Token('"')]) sentence.infer_space_after() assert sentence.to_tokenized_string() == 'xyz " abc "' assert sentence.to_plain_string() == 'xyz "abc"' sentence: Sentence = Sentence('xyz " abc "') sentence.infer_space_after() assert sentence.to_tokenized_string() == 'xyz " abc "' assert sentence.to_plain_string() == 'xyz "abc"' def test_sentence_get_item(): sentence: Sentence = Sentence("I love Berlin.", use_tokenizer=SegtokTokenizer()) assert sentence.get_token(1) == sentence[0] assert sentence.get_token(3) == sentence[2] with pytest.raises(IndexError): _ = sentence[4] def test_token_positions_when_creating_with_tokenizer(): sentence = Sentence("I love Berlin .", use_tokenizer=SpaceTokenizer()) assert sentence.tokens[0].start_position == 0 assert sentence.tokens[0].end_position == 1 assert sentence.tokens[1].start_position == 2 assert sentence.tokens[1].end_position == 6 assert sentence.tokens[2].start_position == 7 assert sentence.tokens[2].end_position == 13 sentence = Sentence(" I love Berlin.", use_tokenizer=SegtokTokenizer()) assert sentence.tokens[0].start_position == 1 assert sentence.tokens[0].end_position == 2 assert sentence.tokens[1].start_position == 3 assert sentence.tokens[1].end_position == 7 assert sentence.tokens[2].start_position == 9 assert sentence.tokens[2].end_position == 15 def test_token_positions_when_creating_word_by_word(): sentence: Sentence = Sentence( [ Token("I"), Token("love"), Token("Berlin"), Token("."), ] ) assert sentence.tokens[0].start_position == 0 assert sentence.tokens[0].end_position == 1 assert sentence.tokens[1].start_position == 2 assert sentence.tokens[1].end_position == 6 assert sentence.tokens[2].start_position == 7 assert sentence.tokens[2].end_position == 13 def no_op_tokenizer(text: str) -> List[str]: return [text]
16,472
34.967249
115
py
flair
flair-master/tests/embedding_test_utils.py
from typing import Any, Dict, List, Optional, Type import pytest import torch from flair.data import Sentence from flair.embeddings import Embeddings from flair.embeddings.base import load_embeddings class BaseEmbeddingsTest: embedding_cls: Type[Embeddings[Sentence]] is_token_embedding: bool is_document_embedding: bool default_args: Dict[str, Any] valid_args: List[Dict[str, Any]] = [] invalid_args: List[Dict[str, Any]] = [] invalid_names: List[str] = [] name_field: Optional[str] = None weired_texts: List[str] = [ "Hybrid mesons , qq ̄ states with an admixture", "typical proportionalities of \u223C 1nmV \u2212 1 [ 3,4 ] .", "🤟 🤟 🤟 hüllo", "🤟hallo 🤟 🤟 🤟 🤟", "🤟", "\uF8F9", ] def create_embedding_from_name(self, name: str): """Overwrite this method if it is more complex to load an embedding by name.""" assert self.name_field is not None kwargs = dict(self.default_args) kwargs.pop(self.name_field) return self.embedding_cls(name, **kwargs) # type: ignore[call-arg] def create_embedding_with_args(self, args: Dict[str, Any]): kwargs = dict(self.default_args) for k, v in args.items(): kwargs[k] = v return self.embedding_cls(**kwargs) @pytest.mark.parametrize("text", weired_texts) def test_embedding_works_with_weird_text(self, text): embeddings = self.create_embedding_with_args(self.default_args) embedding_names = embeddings.get_names() sentence = Sentence(text) embeddings.embed(sentence) if self.is_token_embedding: for token in sentence: assert len(token.get_embedding(embedding_names)) == embeddings.embedding_length if self.is_document_embedding: assert len(sentence.get_embedding(embedding_names)) == embeddings.embedding_length @pytest.mark.parametrize("args", valid_args) def test_embedding_also_sets_trailing_whitespaces(self, args): if not self.is_token_embedding: pytest.skip("The test is only valid for token embeddings") embeddings = self.create_embedding_with_args(args) sentence: Sentence = Sentence(["hello", " ", "hm", " "]) embeddings.embed(sentence) names = embeddings.get_names() for token in sentence: assert len(token.get_embedding(names)) == embeddings.embedding_length @pytest.mark.parametrize("args", valid_args) def test_generic_sentence(self, args): embeddings = self.create_embedding_with_args(args) sentence: Sentence = Sentence("I love Berlin") embeddings.embed(sentence) names = embeddings.get_names() if self.is_token_embedding: for token in sentence: assert len(token.get_embedding(names)) == embeddings.embedding_length if self.is_document_embedding: assert len(sentence.get_embedding(names)) == embeddings.embedding_length @pytest.mark.parametrize("name", invalid_names) def test_load_non_existing_embedding(self, name): with pytest.raises(ValueError): self.create_embedding_from_name(name) def test_keep_batch_order(self): embeddings = self.create_embedding_with_args(self.default_args) embedding_names = embeddings.get_names() sentences_1 = [Sentence("First sentence"), Sentence("This is second sentence")] sentences_2 = [Sentence("This is second sentence"), Sentence("First sentence")] embeddings.embed(sentences_1) embeddings.embed(sentences_2) assert sentences_1[0].to_original_text() == "First sentence" assert sentences_1[1].to_original_text() == "This is second sentence" if self.is_document_embedding: assert ( torch.norm( sentences_1[0].get_embedding(embedding_names) - sentences_2[1].get_embedding(embedding_names) ) == 0.0 ) assert ( torch.norm( sentences_1[1].get_embedding(embedding_names) - sentences_2[0].get_embedding(embedding_names) ) == 0.0 ) if self.is_token_embedding: for i in range(len(sentences_1[0])): assert ( torch.norm( sentences_1[0][i].get_embedding(embedding_names) - sentences_2[1][i].get_embedding(embedding_names) ) == 0.0 ) for i in range(len(sentences_1[1])): assert ( torch.norm( sentences_1[1][i].get_embedding(embedding_names) - sentences_2[0][i].get_embedding(embedding_names) ) == 0.0 ) del embeddings @pytest.mark.parametrize("args", valid_args) def test_embeddings_stay_the_same_after_saving_and_loading(self, args): embeddings = self.create_embedding_with_args(args) sentence_old: Sentence = Sentence("I love Berlin") embeddings.embed(sentence_old) names_old = embeddings.get_names() embedding_length_old = embeddings.embedding_length save_data = embeddings.save_embeddings(use_state_dict=True) del embeddings new_embeddings = load_embeddings(save_data) sentence_new: Sentence = Sentence("I love Berlin") new_embeddings.embed(sentence_new) names_new = new_embeddings.get_names() embedding_length_new = new_embeddings.embedding_length assert names_old == names_new assert embedding_length_old == embedding_length_new if self.is_token_embedding: for token_old, token_new in zip(sentence_old, sentence_new): assert (token_old.get_embedding(names_old) == token_new.get_embedding(names_new)).all() if self.is_document_embedding: assert (sentence_old.get_embedding(names_old) == sentence_new.get_embedding(names_new)).all() def test_default_embeddings_stay_the_same_after_saving_and_loading(self): embeddings = self.create_embedding_with_args(self.default_args) sentence_old: Sentence = Sentence("I love Berlin") embeddings.embed(sentence_old) names_old = embeddings.get_names() embedding_length_old = embeddings.embedding_length save_data = embeddings.save_embeddings(use_state_dict=True) new_embeddings = load_embeddings(save_data) sentence_new: Sentence = Sentence("I love Berlin") new_embeddings.embed(sentence_new) names_new = new_embeddings.get_names() embedding_length_new = new_embeddings.embedding_length assert not new_embeddings.training assert names_old == names_new assert embedding_length_old == embedding_length_new if self.is_token_embedding: for token_old, token_new in zip(sentence_old, sentence_new): assert (token_old.get_embedding(names_old) == token_new.get_embedding(names_new)).all() if self.is_document_embedding: assert (sentence_old.get_embedding(names_old) == sentence_new.get_embedding(names_new)).all() def test_embeddings_load_in_eval_mode(self): embeddings = self.create_embedding_with_args(self.default_args) assert not embeddings.training
7,511
39.387097
113
py
flair
flair-master/tests/test_trainer.py
import pytest from torch.optim import Adam import flair from flair.data import Sentence from flair.datasets import ClassificationCorpus from flair.embeddings import DocumentPoolEmbeddings, FlairEmbeddings, WordEmbeddings from flair.models import SequenceTagger, TextClassifier from flair.trainers import ModelTrainer turian_embeddings = WordEmbeddings("turian") @pytest.mark.integration() def test_text_classifier_multi(results_base_path, tasks_base_path): flair.set_seed(123) flair_embeddings = FlairEmbeddings("news-forward-fast") corpus = ClassificationCorpus( tasks_base_path / "trivial" / "trivial_text_classification_single", label_type="city", ) label_dict = corpus.make_label_dictionary(label_type="city") model: TextClassifier = TextClassifier( embeddings=DocumentPoolEmbeddings([flair_embeddings], fine_tune_mode="linear"), label_dictionary=label_dict, label_type="city", ) trainer = ModelTrainer(model, corpus) trainer.train(results_base_path, mini_batch_size=2, max_epochs=1, shuffle=True) del model train_log_file = results_base_path / "training.log" assert train_log_file.exists() lines = train_log_file.read_text(encoding="utf-8").split("\n") expected_substrings = [ "compute on device: ", "Corpus: ", "- learning_rate: ", "patience", "embedding storage:", "epoch 1 - iter", "EPOCH 1 done: loss", "Results:", ] for expected_substring in expected_substrings: assert any(expected_substring in line for line in lines), expected_substring @pytest.mark.integration() def test_train_load_use_tagger_large(results_base_path, tasks_base_path): corpus = flair.datasets.UD_ENGLISH().downsample(0.01) tag_dictionary = corpus.make_label_dictionary("pos") tagger: SequenceTagger = SequenceTagger( hidden_size=64, embeddings=turian_embeddings, tag_dictionary=tag_dictionary, tag_type="pos", use_crf=False, ) # initialize trainer trainer: ModelTrainer = ModelTrainer(tagger, corpus) trainer.train( results_base_path, learning_rate=0.1, mini_batch_size=32, max_epochs=2, shuffle=False, ) del trainer, tagger, tag_dictionary, corpus loaded_model: SequenceTagger = SequenceTagger.load(results_base_path / "final-model.pt") sentence = Sentence("I love Berlin") sentence_empty = Sentence(" ") loaded_model.predict(sentence) loaded_model.predict([sentence, sentence_empty]) loaded_model.predict([sentence_empty]) del loaded_model @pytest.mark.integration() def test_train_load_use_tagger_adam(results_base_path, tasks_base_path): corpus = flair.datasets.ColumnCorpus(data_folder=tasks_base_path / "fashion", column_format={0: "text", 3: "ner"}) tag_dictionary = corpus.make_label_dictionary("ner", add_unk=False) tagger: SequenceTagger = SequenceTagger( hidden_size=64, embeddings=turian_embeddings, tag_dictionary=tag_dictionary, tag_type="ner", use_crf=False, ) # initialize trainer trainer: ModelTrainer = ModelTrainer(tagger, corpus) trainer.train( results_base_path, learning_rate=0.1, mini_batch_size=2, max_epochs=2, shuffle=False, optimizer=Adam, ) del trainer, tagger, tag_dictionary, corpus loaded_model: SequenceTagger = SequenceTagger.load(results_base_path / "final-model.pt") sentence = Sentence("I love Berlin") sentence_empty = Sentence(" ") loaded_model.predict(sentence) loaded_model.predict([sentence, sentence_empty]) loaded_model.predict([sentence_empty]) del loaded_model def test_missing_validation_split(results_base_path, tasks_base_path): corpus = flair.datasets.ColumnCorpus( data_folder=tasks_base_path / "fewshot_conll", train_file="1shot.txt", sample_missing_splits=False, column_format={0: "text", 1: "ner"}, ) tag_dictionary = corpus.make_label_dictionary("ner", add_unk=True) tagger: SequenceTagger = SequenceTagger( hidden_size=64, embeddings=turian_embeddings, tag_dictionary=tag_dictionary, tag_type="ner", use_crf=False, ) # initialize trainer trainer: ModelTrainer = ModelTrainer(tagger, corpus) trainer.train( results_base_path, learning_rate=0.1, mini_batch_size=2, max_epochs=2, shuffle=False, optimizer=Adam, ) del trainer, tagger, tag_dictionary, corpus loaded_model: SequenceTagger = SequenceTagger.load(results_base_path / "final-model.pt") sentence = Sentence("I love Berlin") sentence_empty = Sentence(" ") loaded_model.predict(sentence) loaded_model.predict([sentence, sentence_empty]) loaded_model.predict([sentence_empty]) del loaded_model
4,999
28.585799
118
py
flair
flair-master/tests/test_language_model.py
import pytest from flair.data import Dictionary, Sentence from flair.embeddings import FlairEmbeddings, TokenEmbeddings from flair.models import LanguageModel from flair.trainers.language_model_trainer import LanguageModelTrainer, TextCorpus @pytest.mark.integration() def test_train_language_model(results_base_path, resources_path): # get default dictionary dictionary: Dictionary = Dictionary.load("chars") # init forward LM with 128 hidden states and 1 layer language_model: LanguageModel = LanguageModel(dictionary, is_forward_lm=True, hidden_size=128, nlayers=1) # get the example corpus and process at character level in forward direction corpus: TextCorpus = TextCorpus( resources_path / "corpora/lorem_ipsum", dictionary, language_model.is_forward_lm, character_level=True, ) # train the language model trainer: LanguageModelTrainer = LanguageModelTrainer(language_model, corpus, test_mode=True) trainer.train(results_base_path, sequence_length=10, mini_batch_size=10, max_epochs=2) # use the character LM as embeddings to embed the example sentence 'I love Berlin' char_lm_embeddings: TokenEmbeddings = FlairEmbeddings(str(results_base_path / "best-lm.pt")) sentence = Sentence("I love Berlin") char_lm_embeddings.embed(sentence) text, likelihood = language_model.generate_text(number_of_characters=100) assert text is not None assert len(text) >= 100 # clean up results directory del trainer, language_model, corpus, char_lm_embeddings @pytest.mark.integration() def test_train_resume_language_model(resources_path, results_base_path, tasks_base_path): # get default dictionary dictionary: Dictionary = Dictionary.load("chars") # init forward LM with 128 hidden states and 1 layer language_model: LanguageModel = LanguageModel(dictionary, is_forward_lm=True, hidden_size=128, nlayers=1) # get the example corpus and process at character level in forward direction corpus: TextCorpus = TextCorpus( resources_path / "corpora/lorem_ipsum", dictionary, language_model.is_forward_lm, character_level=True, ) # train the language model trainer: LanguageModelTrainer = LanguageModelTrainer(language_model, corpus, test_mode=True) trainer.train( results_base_path, sequence_length=10, mini_batch_size=10, max_epochs=2, checkpoint=True, ) del trainer, language_model trainer = LanguageModelTrainer.load_checkpoint(results_base_path / "checkpoint.pt", corpus) trainer.train(results_base_path, sequence_length=10, mini_batch_size=10, max_epochs=2) del trainer def test_generate_text_with_small_temperatures(): from flair.embeddings import FlairEmbeddings language_model = FlairEmbeddings("news-forward-fast", has_decoder=True).lm text, likelihood = language_model.generate_text(temperature=0.01, number_of_characters=100) assert text is not None assert len(text) >= 100 del language_model def test_compute_perplexity(): from flair.embeddings import FlairEmbeddings language_model = FlairEmbeddings("news-forward-fast", has_decoder=True).lm grammatical = "The company made a profit" perplexity_gramamtical_sentence = language_model.calculate_perplexity(grammatical) ungrammatical = "Nook negh qapla!" perplexity_ungramamtical_sentence = language_model.calculate_perplexity(ungrammatical) print(f'"{grammatical}" - perplexity is {perplexity_gramamtical_sentence}') print(f'"{ungrammatical}" - perplexity is {perplexity_ungramamtical_sentence}') assert perplexity_gramamtical_sentence < perplexity_ungramamtical_sentence language_model = FlairEmbeddings("news-backward-fast", has_decoder=True).lm grammatical = "The company made a profit" perplexity_gramamtical_sentence = language_model.calculate_perplexity(grammatical) ungrammatical = "Nook negh qapla!" perplexity_ungramamtical_sentence = language_model.calculate_perplexity(ungrammatical) print(f'"{grammatical}" - perplexity is {perplexity_gramamtical_sentence}') print(f'"{ungrammatical}" - perplexity is {perplexity_ungramamtical_sentence}') assert perplexity_gramamtical_sentence < perplexity_ungramamtical_sentence del language_model
4,360
36.594828
109
py
flair
flair-master/tests/test_tars.py
from flair.data import Sentence from flair.datasets import ClassificationCorpus from flair.models import TARSClassifier from flair.trainers import ModelTrainer def test_init_tars_and_switch(tasks_base_path): # test corpus corpus = ClassificationCorpus(tasks_base_path / "imdb") # create a TARS classifier tars = TARSClassifier( task_name="2_CLASS", label_dictionary=corpus.make_label_dictionary(label_type="class"), label_type="class", ) # check if right number of classes assert len(tars.get_current_label_dictionary()) == 2 # switch to task with only one label tars.add_and_switch_to_new_task("1_CLASS", "one class", "testlabel") # check if right number of classes assert len(tars.get_current_label_dictionary()) == 1 # switch to task with three labels provided as list tars.add_and_switch_to_new_task("3_CLASS", ["list 1", "list 2", "list 3"], "testlabel") # check if right number of classes assert len(tars.get_current_label_dictionary()) == 3 # switch to task with four labels provided as set tars.add_and_switch_to_new_task("4_CLASS", {"set 1", "set 2", "set 3", "set 4"}, "testlabel") # check if right number of classes assert len(tars.get_current_label_dictionary()) == 4 # switch to task with two labels provided as Dictionary tars.add_and_switch_to_new_task("2_CLASS_AGAIN", corpus.make_label_dictionary(label_type="class"), "testlabel") # check if right number of classes assert len(tars.get_current_label_dictionary()) == 2 def test_train_tars(tasks_base_path, results_base_path): # test corpus corpus = ClassificationCorpus(tasks_base_path / "imdb_underscore") # create a TARS classifier tars = TARSClassifier(embeddings="sshleifer/tiny-distilroberta-base") # switch to a new task (TARS can do multiple tasks so you must define one) tars.add_and_switch_to_new_task( task_name="question 2_CLASS", label_dictionary=corpus.make_label_dictionary(label_type="class"), label_type="class", ) # initialize the text classifier trainer trainer = ModelTrainer(tars, corpus) # start the training trainer.train( base_path=results_base_path, learning_rate=0.02, mini_batch_size=1, max_epochs=1, ) sentence = Sentence("This is great!") tars.predict(sentence)
2,404
31.945205
115
py
flair
flair-master/tests/conftest.py
from pathlib import Path import pytest import torch import flair @pytest.fixture(scope="module") def resources_path(): return Path(__file__).parent / "resources" @pytest.fixture(scope="module") def tasks_base_path(resources_path): return resources_path / "tasks" @pytest.fixture() def results_base_path(resources_path): path = resources_path / "results" try: yield path finally: for p in reversed(list(path.rglob("*"))): if p.is_file(): p.unlink() else: p.rmdir() if path.is_dir(): path.rmdir() @pytest.fixture(autouse=True) def set_cpu(force_cpu): if force_cpu: flair.device = torch.device("cpu") def pytest_addoption(parser): parser.addoption( "--runintegration", action="store_true", default=False, help="run integration tests", ) parser.addoption( "--force-cpu", action="store_true", default=False, help="use cpu for tests even when gpu is available", ) def pytest_collection_modifyitems(config, items): if not config.getoption("--runintegration"): skip_integration = pytest.mark.skip(reason="need --runintegration option to run") for item in items: if "integration" in item.keywords: item.add_marker(skip_integration) def pytest_generate_tests(metafunc): option_value = metafunc.config.getoption("--force-cpu") if "force_cpu" in metafunc.fixturenames and option_value is not None: metafunc.parametrize("force_cpu", [option_value])
1,620
23.19403
89
py
flair
flair-master/tests/test_corpus_dictionary.py
import os import pytest import flair from flair.data import Corpus, Dictionary, Label, Sentence from flair.datasets import ColumnCorpus, FlairDatapointDataset, SentenceDataset def test_dictionary_get_items_with_unk(): dictionary: Dictionary = Dictionary(add_unk=True) dictionary.add_item("class_1") dictionary.add_item("class_2") dictionary.add_item("class_3") items = dictionary.get_items() assert len(items) == 4 assert items[0] == "<unk>" assert items[1] == "class_1" assert items[2] == "class_2" assert items[3] == "class_3" def test_dictionary_get_items_without_unk(): dictionary: Dictionary = Dictionary(add_unk=False) dictionary.add_item("class_1") dictionary.add_item("class_2") dictionary.add_item("class_3") items = dictionary.get_items() assert len(items) == 3 assert items[0] == "class_1" assert items[1] == "class_2" assert items[2] == "class_3" def test_dictionary_get_idx_for_item(): dictionary: Dictionary = Dictionary(add_unk=False) dictionary.add_item("class_1") dictionary.add_item("class_2") dictionary.add_item("class_3") idx = dictionary.get_idx_for_item("class_2") assert idx == 1 def test_dictionary_get_item_for_index(): dictionary: Dictionary = Dictionary(add_unk=False) dictionary.add_item("class_1") dictionary.add_item("class_2") dictionary.add_item("class_3") item = dictionary.get_item_for_index(0) assert item == "class_1" def test_dictionary_save_and_load(): dictionary: Dictionary = Dictionary(add_unk=False) dictionary.add_item("class_1") dictionary.add_item("class_2") dictionary.add_item("class_3") file_path = "dictionary.txt" dictionary.save(file_path) loaded_dictionary = dictionary.load_from_file(file_path) assert len(dictionary) == len(loaded_dictionary) assert len(dictionary.get_items()) == len(loaded_dictionary.get_items()) # clean up file os.remove(file_path) def test_deprecated_sentence_dataset(): with pytest.warns(DeprecationWarning): # test to make sure the warning comes, but class works SentenceDataset([Sentence("Short sentences are short")]) def test_tagged_corpus_get_all_sentences(): train_sentence = Sentence("I'm used in training.") dev_sentence = Sentence("I'm a dev sentence.") test_sentence = Sentence("I will be only used for testing.") corpus: Corpus = Corpus( FlairDatapointDataset([train_sentence]), FlairDatapointDataset([dev_sentence]), FlairDatapointDataset([test_sentence]), ) all_sentences = corpus.get_all_sentences() assert len(all_sentences) == 3 def test_tagged_corpus_make_vocab_dictionary(): train_sentence = Sentence("used in training. training is cool.") corpus: Corpus = Corpus(FlairDatapointDataset([train_sentence]), sample_missing_splits=False) vocab = corpus.make_vocab_dictionary(max_tokens=2, min_freq=-1) assert len(vocab) == 3 assert "<unk>" in vocab.get_items() assert "training" in vocab.get_items() assert "." in vocab.get_items() vocab = corpus.make_vocab_dictionary(max_tokens=-1, min_freq=-1) assert len(vocab) == 7 vocab = corpus.make_vocab_dictionary(max_tokens=-1, min_freq=2) assert len(vocab) == 3 assert "<unk>" in vocab.get_items() assert "training" in vocab.get_items() assert "." in vocab.get_items() def test_label_set_confidence(): label = Label(data_point=None, value="class_1", score=3.2) assert label.score == 3.2 assert label.value == "class_1" label._score = 0.2 assert label.score == 0.2 def test_tagged_corpus_make_label_dictionary(): sentence_1 = Sentence("sentence 1").add_label("label", "class_1") sentence_2 = Sentence("sentence 2").add_label("label", "class_2") sentence_3 = Sentence("sentence 3").add_label("label", "class_1") corpus: Corpus = Corpus( FlairDatapointDataset([sentence_1, sentence_2, sentence_3]), FlairDatapointDataset([]), FlairDatapointDataset([]), ) label_dict = corpus.make_label_dictionary("label", add_unk=True) assert len(label_dict) == 3 assert "<unk>" in label_dict.get_items() assert "class_1" in label_dict.get_items() assert "class_2" in label_dict.get_items() with pytest.warns(DeprecationWarning): # test to make sure the warning comes, but function works corpus.make_tag_dictionary("label") def test_obtain_statistics(): sentence_1 = Sentence("The snake hissed to the mountain goat") sentence_1_labels = " O B-Ani O O O B-Ani E-Ani".split() sentence_2 = Sentence("Saber tooth tigers are extinct") sentence_2_labels = " B-Ani I-Ani E-Ani O O".split() for sentence, labels in [(sentence_1, sentence_1_labels), (sentence_2, sentence_2_labels)]: assert len(sentence) == len(labels) for token, label in zip(sentence, labels): token.add_label("ner", label) corpus = Corpus( FlairDatapointDataset([sentence_1, sentence_2]), FlairDatapointDataset([]), FlairDatapointDataset([sentence_2]), ) statistics = corpus.obtain_statistics("ner", pretty_print=False) assert statistics == { "TRAIN": { "dataset": "TRAIN", "total_number_of_documents": 2, "number_of_documents_per_class": {"O": 6, "B-Ani": 3, "E-Ani": 2, "I-Ani": 1}, "number_of_tokens_per_tag": {"O": 6, "B-Ani": 3, "E-Ani": 2, "I-Ani": 1}, "number_of_tokens": {"total": 12, "min": 5, "max": 7, "avg": 6.0}, }, "TEST": { "dataset": "TEST", "total_number_of_documents": 1, "number_of_documents_per_class": {"B-Ani": 1, "I-Ani": 1, "E-Ani": 1, "O": 2}, "number_of_tokens_per_tag": {"B-Ani": 1, "I-Ani": 1, "E-Ani": 1, "O": 2}, "number_of_tokens": {"total": 5, "min": 5, "max": 5, "avg": 5.0}, }, "DEV": {}, } def test_tagged_corpus_statistics(): train_sentence = Sentence("I love Berlin.", use_tokenizer=True).add_label("label", "class_1") dev_sentence = Sentence("The sun is shining.", use_tokenizer=True).add_label("label", "class_2") test_sentence = Sentence("Berlin is sunny.", use_tokenizer=True).add_label("label", "class_1") class_to_count_dict = Corpus._count_sentence_labels([train_sentence, dev_sentence, test_sentence]) assert "class_1" in class_to_count_dict assert "class_2" in class_to_count_dict assert class_to_count_dict["class_1"] == 2 assert class_to_count_dict["class_2"] == 1 tokens_in_sentences = Corpus._get_tokens_per_sentence([train_sentence, dev_sentence, test_sentence]) assert len(tokens_in_sentences) == 3 assert tokens_in_sentences[0] == 4 assert tokens_in_sentences[1] == 5 assert tokens_in_sentences[2] == 4 def test_tagged_corpus_statistics_multi_label(): train_sentence = Sentence("I love Berlin.", use_tokenizer=True).add_label("label", "class_1") dev_sentence = Sentence("The sun is shining.", use_tokenizer=True).add_label("label", "class_2") test_sentence = Sentence("Berlin is sunny.", use_tokenizer=True) test_sentence.add_label("label", "class_1") test_sentence.add_label("label", "class_2") class_to_count_dict = Corpus._count_sentence_labels([train_sentence, dev_sentence, test_sentence]) assert "class_1" in class_to_count_dict assert "class_2" in class_to_count_dict assert class_to_count_dict["class_1"] == 2 assert class_to_count_dict["class_2"] == 2 tokens_in_sentences = Corpus._get_tokens_per_sentence([train_sentence, dev_sentence, test_sentence]) assert len(tokens_in_sentences) == 3 assert tokens_in_sentences[0] == 4 assert tokens_in_sentences[1] == 5 assert tokens_in_sentences[2] == 4 def test_tagged_corpus_downsample(): sentence = Sentence("I love Berlin.", use_tokenizer=True).add_label("label", "class_1") corpus: Corpus = Corpus( FlairDatapointDataset( [ sentence, sentence, sentence, sentence, sentence, sentence, sentence, sentence, sentence, sentence, ] ), sample_missing_splits=False, ) assert len(corpus.train) == 10 corpus.downsample(percentage=0.3, downsample_dev=False, downsample_test=False) assert len(corpus.train) == 3 def test_classification_corpus_multi_labels_without_negative_examples(tasks_base_path): corpus = flair.datasets.ClassificationCorpus( tasks_base_path / "multi_class_negative_examples", allow_examples_without_labels=False, ) assert len(corpus.train) == 7 assert len(corpus.dev) == 4 assert len(corpus.test) == 5 def test_classification_corpus_multi_labels_with_negative_examples(tasks_base_path): corpus = flair.datasets.ClassificationCorpus( tasks_base_path / "multi_class_negative_examples", allow_examples_without_labels=True, ) assert len(corpus.train) == 8 assert len(corpus.dev) == 5 assert len(corpus.test) == 6 def test_misalignment_spans(tasks_base_path): example_txt = """George B-NAME Washington I-NAME went O \t O Washington B-CITY and O enjoyed O some O coffee B-BEVERAGE """ train_path = tasks_base_path / "tmp" / "train.txt" try: train_path.parent.mkdir(exist_ok=True, parents=True) train_path.write_text(example_txt, encoding="utf-8") corpus = ColumnCorpus( data_folder=train_path.parent, column_format={0: "text", 1: "ner"}, train_file=train_path.name ) sentence = corpus.train[0] span_texts = [span.text for span in sentence.get_spans("ner")] assert span_texts == ["George Washington", "Washington", "coffee"] finally: train_path.unlink() train_path.parent.rmdir()
10,036
30.662461
106
py
flair
flair-master/tests/model_test_utils.py
from typing import Any, Dict, List, Optional, Type import pytest import flair from flair.data import Dictionary, Sentence from flair.embeddings import TransformerEmbeddings from flair.models import FewshotClassifier from flair.nn import Model from flair.trainers import ModelTrainer class BaseModelTest: model_cls: Type[Model] pretrained_model: Optional[str] = None empty_sentence = Sentence(" ") train_label_type: str multiclass_prediction_labels: List[str] model_args: Dict[str, Any] = {} training_args: Dict[str, Any] = {} finetune_instead_of_train: bool = False @pytest.fixture() def embeddings(self): pytest.skip("This test requires the `embeddings` fixture to be defined") @pytest.fixture() def corpus(self, tasks_base_path): pytest.skip("This test requires the `corpus` fixture to be defined") @pytest.fixture() def multi_class_corpus(self, tasks_base_path): pytest.skip("This test requires the `multi_class_corpus` fixture to be defined") @pytest.fixture() def multi_corpus(self, tasks_base_path): pytest.skip("This test requires the `multi_corpus` fixture to be defined") @pytest.fixture() def example_sentence(self): return Sentence("I love Berlin") @pytest.fixture() def train_test_sentence(self): return Sentence("Berlin is a really nice city.") @pytest.fixture() def labeled_sentence(self): pytest.skip("This test requires the `labeled_sentence` fixture to be defined") @pytest.fixture() def multiclass_train_test_sentence(self): pytest.skip("This test requires the `multiclass_train_test_sentence` fixture to be defined") def transform_corpus(self, model, corpus): return corpus def assert_training_example(self, predicted_training_example): pass def build_model(self, embeddings, label_dict, **kwargs): model_args = dict(self.model_args) for k in kwargs: if k in model_args: del model_args[k] return self.model_cls( embeddings=embeddings, label_dictionary=label_dict, label_type=self.train_label_type, **model_args, **kwargs, ) def has_embedding(self, sentence): return sentence.get_embedding().cpu().numpy().size > 0 @pytest.fixture() def loaded_pretrained_model(self): if self.pretrained_model is None: pytest.skip("For this test `pretrained_model` needs to be set.") return self.model_cls.load(self.pretrained_model) @pytest.mark.integration() def test_load_use_model(self, example_sentence, loaded_pretrained_model): loaded_pretrained_model.predict(example_sentence) loaded_pretrained_model.predict([example_sentence, self.empty_sentence]) loaded_pretrained_model.predict([self.empty_sentence]) del loaded_pretrained_model example_sentence.clear_embeddings() self.empty_sentence.clear_embeddings() @pytest.mark.integration() def test_train_load_use_model(self, results_base_path, corpus, embeddings, example_sentence, train_test_sentence): flair.set_seed(123) label_dict = corpus.make_label_dictionary(label_type=self.train_label_type) model = self.build_model(embeddings, label_dict) corpus = self.transform_corpus(model, corpus) trainer = ModelTrainer(model, corpus) if self.finetune_instead_of_train: trainer.fine_tune(results_base_path, shuffle=False, **self.training_args) else: trainer.train(results_base_path, shuffle=False, **self.training_args) model.predict(train_test_sentence) self.assert_training_example(train_test_sentence) for label in train_test_sentence.get_labels(self.train_label_type): assert label.value is not None assert 0.0 <= label.score <= 1.0 assert isinstance(label.score, float) del trainer, model, corpus loaded_model = self.model_cls.load(results_base_path / "final-model.pt") loaded_model.predict(example_sentence) loaded_model.predict([example_sentence, self.empty_sentence]) loaded_model.predict([self.empty_sentence]) del loaded_model @pytest.mark.integration() def test_train_load_use_model_multi_corpus( self, results_base_path, multi_corpus, embeddings, example_sentence, train_test_sentence ): flair.set_seed(123) label_dict = multi_corpus.make_label_dictionary(label_type=self.train_label_type) model = self.build_model(embeddings, label_dict) corpus = self.transform_corpus(model, multi_corpus) trainer = ModelTrainer(model, corpus) if self.finetune_instead_of_train: trainer.fine_tune(results_base_path, shuffle=False, **self.training_args) else: trainer.train(results_base_path, shuffle=False, **self.training_args) model.predict(train_test_sentence) self.assert_training_example(train_test_sentence) for label in train_test_sentence.get_labels(self.train_label_type): assert label.value is not None assert 0.0 <= label.score <= 1.0 assert isinstance(label.score, float) del trainer, model, corpus loaded_model = self.model_cls.load(results_base_path / "final-model.pt") loaded_model.predict(example_sentence) loaded_model.predict([example_sentence, self.empty_sentence]) loaded_model.predict([self.empty_sentence]) del loaded_model def test_forward_loss(self, labeled_sentence, embeddings): label_dict = Dictionary() for label in labeled_sentence.get_labels(self.train_label_type): label_dict.add_item(label.value) model = self.build_model(embeddings, label_dict) loss, count = model.forward_loss([labeled_sentence]) assert loss.size() == () assert count == len(labeled_sentence.get_labels(self.train_label_type)) def test_load_use_model_keep_embedding(self, example_sentence, loaded_pretrained_model): assert not self.has_embedding(example_sentence) loaded_pretrained_model.predict(example_sentence, embedding_storage_mode="cpu") assert self.has_embedding(example_sentence) del loaded_pretrained_model def test_train_load_use_model_multi_label( self, results_base_path, multi_class_corpus, embeddings, example_sentence, multiclass_train_test_sentence ): flair.set_seed(123) label_dict = multi_class_corpus.make_label_dictionary(label_type=self.train_label_type) model = self.build_model(embeddings, label_dict, multi_label=True) corpus = self.transform_corpus(model, multi_class_corpus) trainer = ModelTrainer(model, corpus) trainer.train( results_base_path, mini_batch_size=1, max_epochs=5, shuffle=False, train_with_test=True, train_with_dev=True, ) model.predict(multiclass_train_test_sentence) sentence = Sentence("apple tv") model.predict(sentence) for label in self.multiclass_prediction_labels: assert label in [label.value for label in sentence.get_labels(self.train_label_type)], label for label in sentence.labels: print(label) assert label.value is not None assert 0.0 <= label.score <= 1.0 assert type(label.score) is float del trainer, model, multi_class_corpus loaded_model = self.model_cls.load(results_base_path / "final-model.pt") loaded_model.predict(example_sentence) loaded_model.predict([example_sentence, self.empty_sentence]) loaded_model.predict([self.empty_sentence]) def test_context_is_set_correctly(self): sentences = [ Sentence("this is a very very very long sentence"), Sentence("this is a shorter sentence"), Sentence(""), Sentence("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"), Sentence("b"), ] embedding = TransformerEmbeddings("distilbert-base-cased", use_context=True) label_dictionary = Dictionary() model = self.build_model(embedding, label_dictionary) if isinstance(model, FewshotClassifier): model.add_and_switch_to_new_task("test", ["a", "b"], "label") model.predict(sentences) for first, second in zip(sentences[:-1], sentences[1:]): assert first.next_sentence() == second assert first == second.previous_sentence() assert sentences[0].previous_sentence() is None assert sentences[-1].next_sentence() is None
8,882
36.167364
118
py
flair
flair-master/tests/test_visual.py
from flair.data import Sentence, Span, Token from flair.embeddings import FlairEmbeddings from flair.visual import Highlighter from flair.visual.ner_html import HTML_PAGE, PARAGRAPH, TAGGED_ENTITY, render_ner_html from flair.visual.training_curves import Plotter def test_highlighter(resources_path): with (resources_path / "visual/snippet.txt").open() as f: sentences = [x for x in f.read().split("\n") if x] embeddings = FlairEmbeddings("news-forward") features = embeddings.lm.get_representation(sentences[0], "", "").squeeze() Highlighter().highlight_selection( features, sentences[0], n=1000, file_=str(resources_path / "visual/highligh.html"), ) # clean up directory (resources_path / "visual/highligh.html").unlink() def test_plotting_training_curves_and_weights(resources_path): plotter = Plotter() plotter.plot_training_curves(resources_path / "visual/loss.tsv") plotter.plot_weights(resources_path / "visual/weights.txt") # clean up directory (resources_path / "visual/weights.png").unlink() (resources_path / "visual/training.png").unlink() def mock_ner_span(text, tag, start, end): span = Span([]).set_label("class", tag) span.start_pos = start span.end_pos = end span.tokens = [Token(text[start:end])] return span def test_html_rendering(): text = ( "Boris Johnson has been elected new Conservative leader in " "a ballot of party members and will become the " "next UK prime minister. &" ) sentence = Sentence(text) print(sentence[0:2].add_label("ner", "PER")) print(sentence[6:7].add_label("ner", "MISC")) print(sentence[19:20].add_label("ner", "LOC")) colors = { "PER": "#F7FF53", "ORG": "#E8902E", "LOC": "yellow", "MISC": "#4647EB", "O": "#ddd", } actual = render_ner_html([sentence], colors=colors) expected_res = HTML_PAGE.format( text=PARAGRAPH.format( sentence=TAGGED_ENTITY.format(color="#F7FF53", entity="Boris Johnson", label="PER") + " has been elected new " + TAGGED_ENTITY.format(color="#4647EB", entity="Conservative", label="MISC") + " leader in a ballot of party members and will become the next " + TAGGED_ENTITY.format(color="yellow", entity="UK", label="LOC") + " prime minister. &amp;" ), title="Flair", ) assert expected_res == actual
2,506
31.141026
95
py
flair
flair-master/tests/test_datasets_biomedical.py
import inspect import logging import os import tempfile from operator import itemgetter from pathlib import Path from typing import Callable, List, Optional, Type import pytest from tqdm import tqdm import flair from flair.data import Sentence, Token, _iter_dataset from flair.datasets import ColumnCorpus, biomedical from flair.datasets.biomedical import ( CoNLLWriter, Entity, HunerDataset, InternalBioNerDataset, filter_nested_entities, ) from flair.splitter import NoSentenceSplitter, SentenceSplitter from flair.tokenization import SpaceTokenizer logger = logging.getLogger("flair") logger.propagate = True def has_balanced_parantheses(text: str) -> bool: stack = [] opening = ["(", "[", "{"] closing = [")", "]", "}"] for c in text: if c in opening: stack.append(c) elif c in closing: if not stack: return False last_paren = stack.pop() if opening.index(last_paren) != closing.index(c): return False return len(stack) == 0 def gene_predicate(member): return inspect.isclass(member) and "HUNER_GENE_" in str(member) def chemical_predicate(member): return inspect.isclass(member) and "HUNER_CHEMICAL_" in str(member) def disease_predicate(member): return inspect.isclass(member) and "HUNER_DISEASE_" in str(member) def species_predicate(member): return inspect.isclass(member) and "HUNER_SPECIES_" in str(member) def cellline_predicate(member): return inspect.isclass(member) and "HUNER_CELL_LINE_" in str(member) CELLLINE_DATASETS = [ i[1] for i in sorted(inspect.getmembers(biomedical, predicate=cellline_predicate), key=itemgetter(0)) ] CHEMICAL_DATASETS = [ i[1] for i in sorted(inspect.getmembers(biomedical, predicate=chemical_predicate), key=itemgetter(0)) ] DISEASE_DATASETS = [ i[1] for i in sorted(inspect.getmembers(biomedical, predicate=disease_predicate), key=itemgetter(0)) ] GENE_DATASETS = [i[1] for i in sorted(inspect.getmembers(biomedical, predicate=gene_predicate), key=itemgetter(0))] SPECIES_DATASETS = [ i[1] for i in sorted(inspect.getmembers(biomedical, predicate=species_predicate), key=itemgetter(0)) ] ALL_DATASETS = CELLLINE_DATASETS + CHEMICAL_DATASETS + DISEASE_DATASETS + GENE_DATASETS + SPECIES_DATASETS def simple_tokenizer(text: str) -> List[str]: tokens: List[str] = [] word = "" index = -1 for index, char in enumerate(text): if char == " " or char == "-": if len(word) > 0: tokens.append(word) word = "" else: word += char # increment for last token in sentence if not followed by whitespace index += 1 if len(word) > 0: tokens.append(word) return tokens def test_write_to_conll(): text = "This is entity1 entity2 and a long entity3" dataset = InternalBioNerDataset( documents={"1": text}, entities_per_document={ "1": [ Entity((text.find("entity1"), text.find("entity1") + len("entity1")), "E"), Entity((text.find("entity2"), text.find("entity2") + len("entity2")), "E"), Entity( ( text.find("a long entity3"), text.find("a long entity3") + len("a long entity3"), ), "E", ), ] }, ) expected_labeling = [ "This O +", "is O +", "entity1 B-E +", "entity2 B-E +", "and O +", "a B-E +", "long I-E +", "entity3 I-E -", ] assert_conll_writer_output(dataset, expected_labeling) def test_conll_writer_one_token_multiple_entities1(): text = "This is entity1 entity2" dataset = InternalBioNerDataset( documents={"1": text}, entities_per_document={ "1": [ Entity((text.find("entity1"), text.find("entity1") + 2), "E"), Entity((text.find("tity1"), text.find("tity1") + 5), "E"), Entity((text.find("entity2"), text.find("entity2") + len("entity2")), "E"), ] }, ) assert_conll_writer_output(dataset, ["This O +", "is O +", "entity1 B-E +", "entity2 B-E -"]) def test_conll_writer_one_token_multiple_entities2(): text = "This is entity1 entity2" dataset = InternalBioNerDataset( documents={"1": text}, entities_per_document={ "1": [ Entity((text.find("entity1"), text.find("entity1") + 2), "E"), Entity((text.find("tity1"), text.find("tity1") + 5), "E"), ] }, ) assert_conll_writer_output(dataset, ["This O +", "is O +", "entity1 B-E +", "entity2 O -"]) def assert_conll_writer_output( dataset: InternalBioNerDataset, expected_output: List[str], sentence_splitter: Optional[SentenceSplitter] = None, ): fd, outfile_path = tempfile.mkstemp() try: sentence_splitter = sentence_splitter if sentence_splitter else NoSentenceSplitter(tokenizer=SpaceTokenizer()) writer = CoNLLWriter(sentence_splitter=sentence_splitter) writer.write_to_conll(dataset, Path(outfile_path)) with open(outfile_path) as f: contents = [line.strip() for line in f.readlines() if line.strip()] finally: os.close(fd) os.remove(outfile_path) assert contents == expected_output def test_filter_nested_entities(caplog): entities_per_document = { "d0": [Entity((0, 1), "t0"), Entity((2, 3), "t1")], "d1": [Entity((0, 6), "t0"), Entity((2, 3), "t1"), Entity((4, 5), "t2")], "d2": [Entity((0, 3), "t0"), Entity((3, 5), "t1")], "d3": [Entity((0, 3), "t0"), Entity((2, 5), "t1"), Entity((4, 7), "t2")], "d4": [Entity((0, 4), "t0"), Entity((3, 5), "t1")], "d5": [Entity((0, 4), "t0"), Entity((3, 9), "t1")], "d6": [Entity((0, 4), "t0"), Entity((2, 6), "t1")], } target = { "d0": [Entity((0, 1), "t0"), Entity((2, 3), "t1")], "d1": [Entity((2, 3), "t1"), Entity((4, 5), "t2")], "d2": [Entity((0, 3), "t0"), Entity((3, 5), "t1")], "d3": [Entity((0, 3), "t0"), Entity((4, 7), "t2")], "d4": [Entity((0, 4), "t0")], "d5": [Entity((3, 9), "t1")], "d6": [Entity((0, 4), "t0")], } dataset = InternalBioNerDataset(documents={}, entities_per_document=entities_per_document) caplog.set_level(logging.WARNING) filter_nested_entities(dataset) assert "WARNING: Corpus modified by filtering nested entities." in caplog.text for key, entities in dataset.entities_per_document.items(): assert key in target assert len(target[key]) == len(entities) for e1, e2 in zip( sorted(target[key], key=lambda x: str(x)), sorted(entities, key=lambda x: str(x)), ): assert str(e1) == str(e2) def sanity_check_all_corpora(check: Callable[[ColumnCorpus], None]): for _, CorpusType in tqdm(ALL_DATASETS): corpus = CorpusType() check(corpus) @pytest.mark.skip(reason="We skip this test because it's only relevant for development purposes") @pytest.mark.parametrize("CorpusType", ALL_DATASETS) def test_sanity_not_starting_with_minus(CorpusType: Type[ColumnCorpus]): corpus = CorpusType() # type: ignore[call-arg] entities_starting_with_minus = [] for sentence in _iter_dataset(corpus.get_all_sentences()): entities = sentence.get_spans("ner") for entity in entities: if str(entity.tokens[0].text).startswith("-"): entities_starting_with_minus.append(" ".join([t.text for t in entity.tokens])) assert len(entities_starting_with_minus) == 0, "|".join(entities_starting_with_minus) @pytest.mark.parametrize("CorpusType", ALL_DATASETS) @pytest.mark.skip(reason="We skip this test because it's only relevant for development purposes") def test_sanity_no_repeating_Bs(CorpusType: Type[ColumnCorpus]): corpus = CorpusType() # type: ignore[call-arg] longest_repeat_tokens: List[Token] = [] repeat_tokens: List[Token] = [] for sentence in _iter_dataset(corpus.get_all_sentences()): for token in sentence.tokens: if token.get_labels()[0].value.startswith("B") or token.get_labels()[0].value.startswith("S"): repeat_tokens.append(token) else: if len(repeat_tokens) > len(longest_repeat_tokens): longest_repeat_tokens = repeat_tokens repeat_tokens = [] assert len(longest_repeat_tokens) < 4 @pytest.mark.parametrize("CorpusType", ALL_DATASETS) @pytest.mark.skip(reason="We skip this test because it's only relevant for development purposes") def test_sanity_no_long_entities(CorpusType: Type[ColumnCorpus]): corpus = CorpusType() # type: ignore[call-arg] longest_entity: List[str] = [] for sentence in _iter_dataset(corpus.get_all_sentences()): entities = sentence.get_spans("ner") for entity in entities: if len(entity.tokens) > len(longest_entity): longest_entity = [t.text for t in entity.tokens] assert len(longest_entity) < 10, " ".join(longest_entity) @pytest.mark.parametrize("CorpusType", ALL_DATASETS) @pytest.mark.skip(reason="We skip this test because it's only relevant for development purposes") def test_sanity_no_unmatched_parentheses(CorpusType: Type[ColumnCorpus]): corpus = CorpusType() # type: ignore[call-arg] unbalanced_entities = [] for sentence in _iter_dataset(corpus.get_all_sentences()): entities = sentence.get_spans("ner") for entity in entities: entity_text = "".join(t.text for t in entity.tokens) if not has_balanced_parantheses(entity_text): unbalanced_entities.append(entity_text) assert unbalanced_entities == [] @pytest.mark.parametrize("CorpusType", ALL_DATASETS) @pytest.mark.skip(reason="We skip this test because it's only relevant for development purposes") def test_sanity_not_too_many_entities(CorpusType: Type[ColumnCorpus]): corpus = CorpusType() # type: ignore[call-arg] n_entities_per_sentence = [] for sentence in _iter_dataset(corpus.get_all_sentences()): entities = sentence.get_spans("ner") n_entities_per_sentence.append(len(entities)) avg_entities_per_sentence = sum(n_entities_per_sentence) / len(n_entities_per_sentence) assert avg_entities_per_sentence <= 5 @pytest.mark.parametrize("CorpusType", ALL_DATASETS) @pytest.mark.skip(reason="We skip this test because it's only relevant for development purposes") def test_sanity_no_misaligned_entities(CorpusType: Type[HunerDataset]): dataset_name = CorpusType.__class__.__name__.lower() base_path = flair.cache_root / "datasets" data_folder = base_path / dataset_name corpus = CorpusType() internal = corpus.to_internal(data_folder) for doc_id, _doc_text in internal.documents.items(): misaligned_starts = [] misaligned_ends: List[int] = [] entities = internal.entities_per_document[doc_id] entity_starts = [i.char_span.start for i in entities] entity_ends = [i.char_span.stop for i in entities] for start in entity_starts: if start not in entity_starts: misaligned_starts.append(start) for end in entity_ends: if end not in entity_ends: misaligned_starts.append(end) assert len(misaligned_starts) <= len(entities) // 10 assert len(misaligned_ends) <= len(entities) // 10 @pytest.mark.skip(reason="We skip this test because it's only relevant for development purposes") def test_scispacy_tokenization(): from flair.tokenization import SciSpacyTokenizer spacy_tokenizer = SciSpacyTokenizer() sentence = Sentence("HBeAg(+) patients", use_tokenizer=spacy_tokenizer) assert len(sentence) == 5 assert sentence[0].text == "HBeAg" assert sentence[0].start_position == 0 assert sentence[1].text == "(" assert sentence[1].start_position == 5 assert sentence[2].text == "+" assert sentence[2].start_position == 6 assert sentence[3].text == ")" assert sentence[3].start_position == 7 assert sentence[4].text == "patients" assert sentence[4].start_position == 9 sentence = Sentence("HBeAg(+)/HBsAg(+)", use_tokenizer=spacy_tokenizer) assert len(sentence) == 9 assert sentence[0].text == "HBeAg" assert sentence[0].start_position == 0 assert sentence[1].text == "(" assert sentence[1].start_position == 5 assert sentence[2].text == "+" assert sentence[2].start_position == 6 assert sentence[3].text == ")" assert sentence[3].start_position == 7 assert sentence[4].text == "/" assert sentence[4].start_position == 8 assert sentence[5].text == "HBsAg" assert sentence[5].start_position == 9 assert sentence[6].text == "(" assert sentence[6].start_position == 14 assert sentence[7].text == "+" assert sentence[7].start_position == 15 assert sentence[8].text == ")" assert sentence[8].start_position == 16 sentence = Sentence("doxorubicin (DOX)-induced", use_tokenizer=spacy_tokenizer) assert len(sentence) == 5 assert sentence[0].text == "doxorubicin" assert sentence[1].text == "(" assert sentence[2].text == "DOX" assert sentence[3].text == ")" assert sentence[4].text == "-induced"
13,573
34.815303
118
py
flair
flair-master/tests/test_sentence.py
from flair.data import Sentence def test_sentence_context(): # make a sentence and some right context sentence = Sentence("George Washington ging nach Washington.") sentence._next_sentence = Sentence("Das ist eine schöne Stadt.") assert sentence.right_context(1) == [sentence._next_sentence[0]] assert sentence.right_context(10) == sentence._next_sentence.tokens[:10] def test_equality(): assert Sentence("Guten Tag!") != Sentence("Good day!") assert Sentence("Guten Tag!", use_tokenizer=True) != Sentence("Guten Tag!", use_tokenizer=False) # TODO: is this desirable? Or should two sentences with same text be considered same objects? assert Sentence("Guten Tag!") != Sentence("Guten Tag!") def test_token_labeling(): sentence = Sentence("This sentence will be labled") assert sentence.get_labels("ner") == [] assert sentence.get_labels() == [] sentence[2].add_label("ner", "B-promise") sentence[3].add_label("ner", "I-promise") sentence[4].add_label("ner", "I-promise") assert [label.value for label in sentence.get_labels()] == ["B-promise", "I-promise", "I-promise"] assert [token.get_label("ner").value for token in sentence] == ["O", "O", "B-promise", "I-promise", "I-promise"] sentence[1].set_label("ner", "B-object") sentence[1].set_label("ner", "B-subject") assert [label.value for label in sentence.get_labels("ner")] == ["B-subject", "B-promise", "I-promise", "I-promise"] assert [token.get_label("ner").value for token in sentence] == [ "O", "B-subject", "B-promise", "I-promise", "I-promise", ] sentence.set_label("class", "positive") sentence.remove_labels("ner") assert sentence.get_labels("ner") == [] assert [label.value for label in sentence.get_labels()] == ["positive"] sentence[0].add_label("pos", "first") sentence[0].add_label("pos", "primero") sentence[0].add_label("pos", "erstes") assert [label.value for label in sentence.get_labels("pos")] == ["first", "primero", "erstes"] assert sentence[0].get_label("pos").value == "first" def test_start_end_position_untokenized() -> None: sentence: Sentence = Sentence("This is a sentence.", start_position=10) assert sentence.start_position == 10 assert sentence.end_position == 29 assert [(token.start_position, token.end_position) for token in sentence] == [ (0, 4), (5, 7), (8, 9), (10, 18), (18, 19), ] def test_start_end_position_pretokenized() -> None: # Initializing a Sentence this way assumes that there is a space after each token sentence: Sentence = Sentence(["This", "is", "a", "sentence", "."], start_position=10) assert sentence.start_position == 10 assert sentence.end_position == 30 assert [(token.start_position, token.end_position) for token in sentence] == [ (0, 4), (5, 7), (8, 9), (10, 18), (19, 20), ]
2,992
38.381579
120
py
flair
flair-master/tests/__init__.py
0
0
0
py
flair
flair-master/tests/test_multitask.py
import pytest import flair from flair.data import Sentence from flair.datasets import SENTEVAL_CR, SENTEVAL_SST_GRANULAR from flair.embeddings import TransformerDocumentEmbeddings from flair.models import MultitaskModel, TextClassifier from flair.nn.multitask import make_multitask_model_and_corpus from flair.trainers import ModelTrainer @pytest.mark.integration() def test_train_load_use_classifier(results_base_path, tasks_base_path): # --- Embeddings that are shared by both models --- # shared_embedding = TransformerDocumentEmbeddings("distilbert-base-uncased", fine_tune=True) # --- Task 1: Sentiment Analysis (5-class) --- # flair.set_seed(123) # Define corpus and model corpus_1 = SENTEVAL_SST_GRANULAR().downsample(0.01) model_1 = TextClassifier( shared_embedding, label_dictionary=corpus_1.make_label_dictionary("class", add_unk=False), label_type="class" ) # -- Task 2: Binary Sentiment Analysis on Customer Reviews -- # flair.set_seed(123) # Define corpus and model corpus_2 = SENTEVAL_CR().downsample(0.01) model_2 = TextClassifier( shared_embedding, label_dictionary=corpus_2.make_label_dictionary("sentiment", add_unk=False), label_type="sentiment", inverse_model=True, ) # -- Define mapping (which tagger should train on which model) -- # multitask_model, multicorpus = make_multitask_model_and_corpus( [ (model_1, corpus_1), (model_2, corpus_2), ] ) # -- Create model trainer and train -- # trainer = ModelTrainer(multitask_model, multicorpus) trainer.fine_tune(results_base_path, max_epochs=1) del trainer, multitask_model, corpus_1, corpus_2 loaded_model = MultitaskModel.load(results_base_path / "final-model.pt") sentence = Sentence("I love Berlin") sentence_empty = Sentence(" ") loaded_model.predict(sentence) loaded_model.predict([sentence, sentence_empty]) loaded_model.predict([sentence_empty]) for label in sentence.labels: assert label.value is not None assert 0.0 <= label.score <= 1.0 assert type(label.score) is float del loaded_model
2,209
31.5
117
py
flair
flair-master/tests/test_utils.py
from flair.data import Dictionary from flair.training_utils import convert_labels_to_one_hot def test_convert_labels_to_one_hot(): label_dict = Dictionary(add_unk=False) label_dict.add_item("class-1") label_dict.add_item("class-2") label_dict.add_item("class-3") one_hot = convert_labels_to_one_hot([["class-2"]], label_dict) assert one_hot[0][0] == 0 assert one_hot[0][1] == 1 assert one_hot[0][2] == 0
440
26.5625
66
py
flair
flair-master/tests/models/test_relation_classifier.py
from operator import itemgetter from typing import Dict, List, Optional, Set, Tuple import pytest from torch.utils.data import Dataset from flair.data import Relation, Sentence from flair.datasets import ColumnCorpus, DataLoader from flair.embeddings import TransformerDocumentEmbeddings from flair.models import RelationClassifier from flair.models.relation_classifier_model import ( EncodedSentence, EncodingStrategy, EntityMarker, EntityMarkerPunct, EntityMask, TypedEntityMarker, TypedEntityMarkerPunct, TypedEntityMask, ) from tests.model_test_utils import BaseModelTest encoding_strategies: Dict[EncodingStrategy, List[Tuple[str, str]]] = { EntityMask(): [("[HEAD]", "[TAIL]") for _ in range(7)], TypedEntityMask(): [ ("[HEAD-ORG]", "[TAIL-PER]"), ("[HEAD-ORG]", "[TAIL-PER]"), ("[HEAD-ORG]", "[TAIL-PER]"), ("[HEAD-LOC]", "[TAIL-PER]"), ("[HEAD-LOC]", "[TAIL-PER]"), ("[HEAD-LOC]", "[TAIL-PER]"), ("[HEAD-ORG]", "[TAIL-PER]"), ], EntityMarker(): [ ("[HEAD] Google [/HEAD]", "[TAIL] Larry Page [/TAIL]"), ("[HEAD] Google [/HEAD]", "[TAIL] Sergey Brin [/TAIL]"), ("[HEAD] Microsoft [/HEAD]", "[TAIL] Bill Gates [/TAIL]"), ("[HEAD] Berlin [/HEAD]", "[TAIL] Konrad Zuse [/TAIL]"), ("[HEAD] Berlin [/HEAD]", "[TAIL] Joseph Weizenbaum [/TAIL]"), ("[HEAD] Germany [/HEAD]", "[TAIL] Joseph Weizenbaum [/TAIL]"), ("[HEAD] MIT [/HEAD]", "[TAIL] Joseph Weizenbaum [/TAIL]"), ], TypedEntityMarker(): [ ("[HEAD-ORG] Google [/HEAD-ORG]", "[TAIL-PER] Larry Page [/TAIL-PER]"), ("[HEAD-ORG] Google [/HEAD-ORG]", "[TAIL-PER] Sergey Brin [/TAIL-PER]"), ("[HEAD-ORG] Microsoft [/HEAD-ORG]", "[TAIL-PER] Bill Gates [/TAIL-PER]"), ("[HEAD-LOC] Berlin [/HEAD-LOC]", "[TAIL-PER] Konrad Zuse [/TAIL-PER]"), ("[HEAD-LOC] Berlin [/HEAD-LOC]", "[TAIL-PER] Joseph Weizenbaum [/TAIL-PER]"), ("[HEAD-LOC] Germany [/HEAD-LOC]", "[TAIL-PER] Joseph Weizenbaum [/TAIL-PER]"), ("[HEAD-ORG] MIT [/HEAD-ORG]", "[TAIL-PER] Joseph Weizenbaum [/TAIL-PER]"), ], EntityMarkerPunct(): [ ("@ Google @", "# Larry Page #"), ("@ Google @", "# Sergey Brin #"), ("@ Microsoft @", "# Bill Gates #"), ("@ Berlin @", "# Konrad Zuse #"), ("@ Berlin @", "# Joseph Weizenbaum #"), ("@ Germany @", "# Joseph Weizenbaum #"), ("@ MIT @", "# Joseph Weizenbaum #"), ], TypedEntityMarkerPunct(): [ ("@ * ORG * Google @", "# ^ PER ^ Larry Page #"), ("@ * ORG * Google @", "# ^ PER ^ Sergey Brin #"), ("@ * ORG * Microsoft @", "# ^ PER ^ Bill Gates #"), ("@ * LOC * Berlin @", "# ^ PER ^ Konrad Zuse #"), ("@ * LOC * Berlin @", "# ^ PER ^ Joseph Weizenbaum #"), ("@ * LOC * Germany @", "# ^ PER ^ Joseph Weizenbaum #"), ("@ * ORG * MIT @", "# ^ PER ^ Joseph Weizenbaum #"), ], } class TestRelationClassifier(BaseModelTest): model_cls = RelationClassifier train_label_type = "relation" multiclass_prediction_labels = ["apple", "tv"] model_args = { "entity_label_types": "ner", "entity_pair_labels": { # Define valid entity pair combinations, used as relation candidates ("ORG", "PER"), # founded_by ("LOC", "PER"), # place_of_birth }, "allow_unk_tag": False, } training_args = {"max_epochs": 2, "learning_rate": 4e-4, "mini_batch_size": 4} finetune_instead_of_train = True @pytest.fixture() def corpus(self, tasks_base_path): return ColumnCorpus( data_folder=tasks_base_path / "conllu", train_file="train.conllup", dev_file="train.conllup", test_file="train.conllup", column_format={1: "text", 2: "pos", 3: "ner"}, ) @pytest.fixture() def embeddings(self): return TransformerDocumentEmbeddings(model="distilbert-base-uncased", layers="-1", fine_tune=True) def transform_corpus(self, model, corpus): return model.transform_corpus(corpus) @pytest.fixture() def example_sentence(self): sentence = Sentence(["Microsoft", "was", "found", "by", "Bill", "Gates"]) sentence[:1].add_label(typename="ner", value="ORG", score=1.0) sentence[4:].add_label(typename="ner", value="PER", score=1.0) return sentence @pytest.fixture() def train_test_sentence(self): sentence: Sentence = Sentence( [ "Intel", "was", "founded", "on", "July", "18", ",", "1968", ",", "by", "semiconductor", "pioneers", "Gordon", "Moore", "and", "Robert", "Noyce", ".", ] ) sentence[:1].add_label(typename="ner", value="ORG", score=1.0) # Intel -> ORG sentence[12:14].add_label(typename="ner", value="PER", score=1.0) # Gordon Moore -> PER sentence[15:17].add_label(typename="ner", value="PER", score=1.0) # Robert Noyce -> PER return sentence def assert_training_example(self, predicted_training_example): relations: List[Relation] = predicted_training_example.get_relations("relation") assert len(relations) == 2 # Intel ----founded_by---> Gordon Moore assert [label.value for label in relations[0].labels] == ["founded_by"] assert ( relations[0].unlabeled_identifier == Relation( first=predicted_training_example[:1], second=predicted_training_example[12:14] ).unlabeled_identifier ) # Intel ----founded_by---> Robert Noyce assert [label.value for label in relations[1].labels] == ["founded_by"] assert ( relations[1].unlabeled_identifier == Relation( first=predicted_training_example[:1], second=predicted_training_example[15:17] ).unlabeled_identifier ) @staticmethod def check_transformation_correctness( split: Optional[Dataset], ground_truth: Set[Tuple[str, Tuple[str, ...]]], ) -> None: # Ground truth is a set of tuples of (<Sentence Text>, <Relation Label Values>) assert split is not None data_loader = DataLoader(split, batch_size=1) assert all(isinstance(sentence, EncodedSentence) for sentence in map(itemgetter(0), data_loader)) assert { (sentence.to_tokenized_string(), tuple(label.value for label in sentence.get_labels("relation"))) for sentence in map(itemgetter(0), data_loader) } == ground_truth @pytest.mark.parametrize( "cross_augmentation", [True, False], ids=["with_cross_augmentation", "without_cross_augmentation"] ) @pytest.mark.parametrize( ("encoding_strategy", "encoded_entity_pairs"), encoding_strategies.items(), ids=[type(encoding_strategy).__name__ for encoding_strategy in encoding_strategies], ) def test_transform_corpus( self, corpus: ColumnCorpus, embeddings: TransformerDocumentEmbeddings, cross_augmentation: bool, encoding_strategy: EncodingStrategy, encoded_entity_pairs: List[Tuple[str, str]], ) -> None: label_dictionary = corpus.make_label_dictionary("relation") model: RelationClassifier = self.build_model( embeddings, label_dictionary, cross_augmentation=cross_augmentation, encoding_strategy=encoding_strategy ) transformed_corpus = model.transform_corpus(corpus) # Check sentence masking and relation label annotation on # training, validation and test dataset (in this test the splits are the same) ground_truth: Set[Tuple[str, Tuple[str, ...]]] = { # Entity pair permutations of: "Larry Page and Sergey Brin founded Google ." (f"{encoded_entity_pairs[0][1]} and Sergey Brin founded {encoded_entity_pairs[0][0]} .", ("founded_by",)), (f"Larry Page and {encoded_entity_pairs[1][1]} founded {encoded_entity_pairs[1][0]} .", ("founded_by",)), # Entity pair permutations of: "Microsoft was founded by Bill Gates ." (f"{encoded_entity_pairs[2][0]} was founded by {encoded_entity_pairs[2][1]} .", ("founded_by",)), # Entity pair permutations of: "Konrad Zuse was born in Berlin on 22 June 1910 ." ( f"{encoded_entity_pairs[3][1]} was born in {encoded_entity_pairs[3][0]} on 22 June 1910 .", ("place_of_birth",), ), # Entity pair permutations of: "Joseph Weizenbaum , a professor at MIT , was born in Berlin , Germany." ( f"{encoded_entity_pairs[4][1]} , a professor at MIT , " f"was born in {encoded_entity_pairs[4][0]} , Germany .", ("place_of_birth",), ), ( f"{encoded_entity_pairs[5][1]} , a professor at MIT , " f"was born in Berlin , {encoded_entity_pairs[5][0]} .", ("place_of_birth",), ), } if cross_augmentation: # This sentence is only included if we transform the corpus with cross augmentation ground_truth.add( ( f"{encoded_entity_pairs[6][1]} , a professor at {encoded_entity_pairs[6][0]} , " f"was born in Berlin , Germany .", ("O",), ) ) for split in (transformed_corpus.train, transformed_corpus.dev, transformed_corpus.test): self.check_transformation_correctness(split, ground_truth)
9,948
40.627615
118
py
flair
flair-master/tests/models/test_sequence_tagger.py
import pytest import flair from flair.embeddings import FlairEmbeddings, WordEmbeddings from flair.models import SequenceTagger from flair.trainers import ModelTrainer from tests.model_test_utils import BaseModelTest class TestSequenceTagger(BaseModelTest): model_cls = SequenceTagger pretrained_model = "ner-fast" train_label_type = "ner" training_args = { "max_epochs": 2, "learning_rate": 0.1, "mini_batch_size": 2, } model_args = { "hidden_size": 64, "use_crf": False, } def has_embedding(self, sentence): return all(token.get_embedding().cpu().numpy().size != 0 for token in sentence) def build_model(self, embeddings, label_dict, **kwargs): model_args = dict(self.model_args) for k in kwargs: if k in model_args: del model_args[k] return self.model_cls( embeddings=embeddings, tag_dictionary=label_dict, tag_type=self.train_label_type, **model_args, **kwargs, ) @pytest.fixture() def embeddings(self): return WordEmbeddings("turian") @pytest.fixture() def corpus(self, tasks_base_path): return flair.datasets.ColumnCorpus(data_folder=tasks_base_path / "fashion", column_format={0: "text", 3: "ner"}) @pytest.mark.integration() def test_all_tag_proba_embedding(self, example_sentence, loaded_pretrained_model): loaded_pretrained_model.predict(example_sentence, return_probabilities_for_all_classes=True) for token in example_sentence: assert len(token.get_tags_proba_dist(loaded_pretrained_model.label_type)) == len( loaded_pretrained_model.label_dictionary ) score_sum = 0.0 for label in token.get_tags_proba_dist(loaded_pretrained_model.label_type): assert label.data_point == token score_sum += label.score assert abs(score_sum - 1.0) < 1.0e-5 @pytest.mark.integration() def test_force_token_predictions(self, example_sentence, loaded_pretrained_model): loaded_pretrained_model.predict(example_sentence, force_token_predictions=True) assert example_sentence.get_token(3).text == "Berlin" assert example_sentence.get_token(3).tag == "S-LOC" @pytest.mark.integration() def test_train_load_use_tagger_flair_embeddings(self, results_base_path, corpus, example_sentence): tag_dictionary = corpus.make_label_dictionary("ner", add_unk=False) model = self.build_model(FlairEmbeddings("news-forward-fast"), tag_dictionary) trainer = ModelTrainer(model, corpus) trainer.train(results_base_path, shuffle=False, **self.training_args) del trainer, model, tag_dictionary, corpus loaded_model = self.model_cls.load(results_base_path / "final-model.pt") loaded_model.predict(example_sentence) loaded_model.predict([example_sentence, self.empty_sentence]) loaded_model.predict([self.empty_sentence]) del loaded_model @pytest.mark.integration() def test_train_load_use_tagger_with_trainable_hidden_state( self, embeddings, results_base_path, corpus, example_sentence ): tag_dictionary = corpus.make_label_dictionary("ner", add_unk=False) model = self.build_model(embeddings, tag_dictionary, train_initial_hidden_state=True) trainer = ModelTrainer(model, corpus) trainer.train(results_base_path, shuffle=False, **self.training_args) del trainer, model, tag_dictionary, corpus loaded_model = self.model_cls.load(results_base_path / "final-model.pt") loaded_model.predict(example_sentence) loaded_model.predict([example_sentence, self.empty_sentence]) loaded_model.predict([self.empty_sentence]) del loaded_model @pytest.mark.integration() def test_train_load_use_tagger_disjunct_tags( self, results_base_path, tasks_base_path, embeddings, example_sentence ): corpus = flair.datasets.ColumnCorpus( data_folder=tasks_base_path / "fashion_disjunct", column_format={0: "text", 3: "ner"}, ) tag_dictionary = corpus.make_label_dictionary("ner", add_unk=True) model = self.build_model(embeddings, tag_dictionary, allow_unk_predictions=True) trainer = ModelTrainer(model, corpus) trainer.train(results_base_path, shuffle=False, **self.training_args) del trainer, model, tag_dictionary, corpus loaded_model = self.model_cls.load(results_base_path / "final-model.pt") loaded_model.predict(example_sentence) loaded_model.predict([example_sentence, self.empty_sentence]) loaded_model.predict([self.empty_sentence]) del loaded_model
4,859
38.193548
120
py
flair
flair-master/tests/models/test_relation_extractor.py
import pytest from flair.data import Sentence from flair.datasets import ColumnCorpus from flair.embeddings import TransformerWordEmbeddings from flair.models import RelationExtractor from tests.model_test_utils import BaseModelTest class TestRelationExtractor(BaseModelTest): model_cls = RelationExtractor train_label_type = "relation" pretrained_model = "relations" model_args = { "entity_label_type": "ner", "train_on_gold_pairs_only": True, "entity_pair_filters": { # Define valid entity pair combinations, used as relation candidates ("ORG", "PER"), # founded_by ("LOC", "PER"), # place_of_birth }, } training_args = { "max_epochs": 4, "mini_batch_size": 2, "learning_rate": 0.1, } @pytest.fixture() def corpus(self, tasks_base_path): return ColumnCorpus( data_folder=tasks_base_path / "conllu", train_file="train.conllup", dev_file="train.conllup", test_file="train.conllup", column_format={1: "text", 2: "pos", 3: "ner"}, ) @pytest.fixture() def example_sentence(self): sentence = Sentence(["Microsoft", "was", "found", "by", "Bill", "Gates"]) sentence[:1].add_label(typename="ner", value="ORG", score=1.0) sentence[4:].add_label(typename="ner", value="PER", score=1.0) return sentence @pytest.fixture() def train_test_sentence(self): sentence = Sentence(["Apple", "was", "founded", "by", "Steve", "Jobs", "."]) sentence[0:1].add_label("ner", "ORG") sentence[4:6].add_label("ner", "PER") return sentence @pytest.fixture() def embeddings(self): return TransformerWordEmbeddings(model="distilbert-base-uncased", fine_tune=True) def assert_training_example(self, predicted_training_example): relations = predicted_training_example.get_relations("relation") assert len(relations) == 1 assert relations[0].tag == "founded_by" def has_embedding(self, sentence): return all(token.get_embedding().cpu().numpy().size != 0 for token in sentence)
2,183
33.666667
102
py
flair
flair-master/tests/models/test_tars_classifier.py
import pytest from flair.data import Sentence from flair.datasets import ClassificationCorpus from flair.embeddings import TransformerDocumentEmbeddings from flair.models import TARSClassifier from tests.model_test_utils import BaseModelTest class TestTarsClassifier(BaseModelTest): model_cls = TARSClassifier train_label_type = "class" model_args = {"task_name": "2_CLASS"} training_args = {"mini_batch_size": 1, "max_epochs": 2} pretrained_model = "tars-base" @pytest.fixture() def corpus(self, tasks_base_path): return ClassificationCorpus(tasks_base_path / "imdb_underscore") @pytest.fixture() def embeddings(self): return TransformerDocumentEmbeddings("distilbert-base-uncased") @pytest.fixture() def example_sentence(self): return Sentence("This is great!") def build_model(self, embeddings, label_dict, **kwargs): model_args = dict(self.model_args) for k in kwargs: if k in model_args: del model_args[k] return self.model_cls( embeddings=embeddings, label_type=self.train_label_type, **model_args, **kwargs, ) def transform_corpus(self, model, corpus): model.add_and_switch_to_new_task( task_name="2_CLASS", label_dictionary=corpus.make_label_dictionary(self.train_label_type), label_type=self.train_label_type, ) return corpus @pytest.mark.integration() def test_predict_zero_shot(self, loaded_pretrained_model): sentence = Sentence("I am so glad you liked it!") loaded_pretrained_model.predict_zero_shot(sentence, ["happy", "sad"]) assert len(sentence.get_labels(loaded_pretrained_model.label_type)) == 1 assert sentence.get_labels(loaded_pretrained_model.label_type)[0].value == "happy" @pytest.mark.integration() def test_predict_zero_shot_single_label_always_predicts(self, loaded_pretrained_model): sentence = Sentence("I hate it") loaded_pretrained_model.predict_zero_shot(sentence, ["happy", "sad"]) # Ensure this is an example that predicts no classes in multilabel assert len(sentence.get_labels(loaded_pretrained_model.label_type)) == 0 loaded_pretrained_model.predict_zero_shot(sentence, ["happy", "sad"], multi_label=False) assert len(sentence.get_labels(loaded_pretrained_model.label_type)) == 1 assert sentence.get_labels(loaded_pretrained_model.label_type)[0].value == "sad" @pytest.mark.integration() def test_init_tars_and_switch(self, tasks_base_path, corpus): tars = TARSClassifier( task_name="2_CLASS", label_dictionary=corpus.make_label_dictionary(label_type="class"), label_type="class", ) # check if right number of classes assert len(tars.get_current_label_dictionary()) == 2 # switch to task with only one label tars.add_and_switch_to_new_task("1_CLASS", "one class", "testlabel") # check if right number of classes assert len(tars.get_current_label_dictionary()) == 1 # switch to task with three labels provided as list tars.add_and_switch_to_new_task("3_CLASS", ["list 1", "list 2", "list 3"], "testlabel") # check if right number of classes assert len(tars.get_current_label_dictionary()) == 3 # switch to task with four labels provided as set tars.add_and_switch_to_new_task("4_CLASS", {"set 1", "set 2", "set 3", "set 4"}, "testlabel") # check if right number of classes assert len(tars.get_current_label_dictionary()) == 4 # switch to task with two labels provided as Dictionary tars.add_and_switch_to_new_task("2_CLASS_AGAIN", corpus.make_label_dictionary(label_type="class"), "testlabel") # check if right number of classes assert len(tars.get_current_label_dictionary()) == 2 @pytest.mark.skip("embeddings are not supported in tars") def test_load_use_model_keep_embedding(self): pass @pytest.mark.skip("tars needs additional setup after loading") def test_load_use_model(self): pass
4,232
38.194444
119
py
flair
flair-master/tests/models/test_text_regressor.py
import pytest import flair from flair.embeddings import DocumentRNNEmbeddings, WordEmbeddings from flair.models.text_regression_model import TextRegressor from tests.model_test_utils import BaseModelTest class TestTextRegressor(BaseModelTest): model_cls = TextRegressor train_label_type = "regression" training_args = { "max_epochs": 3, "mini_batch_size": 2, "learning_rate": 0.1, "main_evaluation_metric": ("correlation", "pearson"), } def build_model(self, embeddings, label_dict, **kwargs): # no need for label_dict return self.model_cls(embeddings, self.train_label_type) @pytest.fixture() def embeddings(self): glove_embedding = WordEmbeddings("turian") return DocumentRNNEmbeddings([glove_embedding], 128, 1, False, 64, False, False) @pytest.fixture() def corpus(self, tasks_base_path): return flair.datasets.ClassificationCorpus(tasks_base_path / "regression", label_type=self.train_label_type)
1,016
31.806452
116
py
flair
flair-master/tests/models/test_entity_linker.py
import pytest from flair.data import Sentence from flair.datasets import NEL_ENGLISH_AIDA from flair.embeddings import TransformerWordEmbeddings from flair.models import EntityLinker from tests.model_test_utils import BaseModelTest class TestEntityLinker(BaseModelTest): model_cls = EntityLinker train_label_type = "nel" training_args = {"max_epochs": 2} @pytest.fixture() def embeddings(self): return TransformerWordEmbeddings(model="distilbert-base-uncased", layers="-1", fine_tune=True) @pytest.fixture() def corpus(self, tasks_base_path): return NEL_ENGLISH_AIDA().downsample(0.01) @pytest.fixture() def train_test_sentence(self): sentence = Sentence("I love NYC and hate OYC") sentence[2:3].add_label("nel", "New York City") sentence[5:6].add_label("nel", "Old York City") return sentence @pytest.fixture() def labeled_sentence(self): sentence = Sentence("I love NYC and hate OYC") sentence[2:3].add_label("nel", "New York City") sentence[5:6].add_label("nel", "Old York City") return sentence
1,132
28.815789
102
py
flair
flair-master/tests/models/__init__.py
0
0
0
py
flair
flair-master/tests/models/test_tars_ner.py
import pytest import flair from flair.data import Sentence from flair.embeddings import TransformerWordEmbeddings from flair.models import TARSTagger from tests.model_test_utils import BaseModelTest class TestTarsTagger(BaseModelTest): model_cls = TARSTagger train_label_type = "ner" model_args = {"task_name": "2_NER"} training_args = {"mini_batch_size": 1, "max_epochs": 2} pretrained_model = "tars-ner" @pytest.fixture() def corpus(self, tasks_base_path): return flair.datasets.ColumnCorpus(data_folder=tasks_base_path / "fashion", column_format={0: "text", 3: "ner"}) @pytest.fixture() def embeddings(self): return TransformerWordEmbeddings("distilbert-base-uncased") @pytest.fixture() def example_sentence(self): return Sentence("George Washington was born in Washington") def build_model(self, embeddings, label_dict, **kwargs): model_args = dict(self.model_args) for k in kwargs: if k in model_args: del model_args[k] return self.model_cls( embeddings=embeddings, label_type=self.train_label_type, **model_args, **kwargs, ) def transform_corpus(self, model, corpus): model.add_and_switch_to_new_task( task_name="2_NER", label_dictionary=corpus.make_label_dictionary(self.train_label_type), label_type=self.train_label_type, ) return corpus @pytest.mark.integration() def test_predict_zero_shot(self, loaded_pretrained_model): sentence = Sentence("George Washington was born in Washington") loaded_pretrained_model.predict_zero_shot(sentence, ["location", "person"]) assert len(sentence.get_labels("location-person")) == 2 assert sorted([label.value for label in sentence.get_labels("location-person")]) == [ "location", "person", ] @pytest.mark.integration() def test_init_tars_and_switch(self, tasks_base_path, corpus): tars = TARSTagger( task_name="2_NER", label_dictionary=corpus.make_label_dictionary(label_type="ner"), label_type="ner", ) # check if right number of classes assert len(tars.get_current_label_dictionary()) == 10 # switch to task with only one label tars.add_and_switch_to_new_task("1_CLASS", "one class", "testlabel") # check if right number of classes assert len(tars.get_current_label_dictionary()) == 1 # switch to task with three labels provided as list tars.add_and_switch_to_new_task("3_CLASS", ["list 1", "list 2", "list 3"], "testlabel") # check if right number of classes assert len(tars.get_current_label_dictionary()) == 3 # switch to task with four labels provided as set tars.add_and_switch_to_new_task("4_CLASS", {"set 1", "set 2", "set 3", "set 4"}, "testlabel") # check if right number of classes assert len(tars.get_current_label_dictionary()) == 4 # switch to task with two labels provided as Dictionary tars.add_and_switch_to_new_task("2_CLASS_AGAIN", corpus.make_label_dictionary(label_type="ner"), "testlabel") # check if right number of classes assert len(tars.get_current_label_dictionary()) == 10 @pytest.mark.skip("embeddings are not supported in tars") def test_load_use_model_keep_embedding(self): pass @pytest.mark.skip("tars needs additional setup after loading") def test_load_use_model(self): pass
3,630
34.950495
120
py
flair
flair-master/tests/models/test_text_classifier.py
import pytest import flair.datasets from flair.data import Sentence from flair.embeddings import DocumentRNNEmbeddings, FlairEmbeddings, WordEmbeddings from flair.models import TextClassifier from flair.samplers import ImbalancedClassificationDatasetSampler from flair.trainers import ModelTrainer from tests.model_test_utils import BaseModelTest class TestTextClassifier(BaseModelTest): model_cls = TextClassifier pretrained_model = "sentiment" train_label_type = "topic" multiclass_prediction_labels = ["apple", "tv"] training_args = { "max_epochs": 4, } @pytest.fixture() def embeddings(self): turian_embeddings = WordEmbeddings("turian") document_embeddings = DocumentRNNEmbeddings([turian_embeddings], 128, 1, False, 64, False, False) return document_embeddings @pytest.fixture() def corpus(self, tasks_base_path): return flair.datasets.ClassificationCorpus(tasks_base_path / "imdb", label_type="topic") @pytest.fixture() def multiclass_train_test_sentence(self): return Sentence("apple tv") @pytest.fixture() def multi_class_corpus(self, tasks_base_path): return flair.datasets.ClassificationCorpus(tasks_base_path / "multi_class", label_type="topic") @pytest.mark.integration() def test_train_load_use_classifier_with_sampler( self, results_base_path, corpus, embeddings, example_sentence, train_test_sentence ): flair.set_seed(123) label_dict = corpus.make_label_dictionary(label_type=self.train_label_type) model = self.model_cls(embeddings=embeddings, label_dictionary=label_dict, label_type=self.train_label_type) trainer = ModelTrainer(model, corpus) trainer.train(results_base_path, max_epochs=2, shuffle=False, sampler=ImbalancedClassificationDatasetSampler) model.predict(train_test_sentence) for label in train_test_sentence.get_labels(self.train_label_type): assert label.value is not None assert 0.0 <= label.score <= 1.0 assert isinstance(label.score, float) del trainer, model, corpus loaded_model = self.model_cls.load(results_base_path / "final-model.pt") loaded_model.predict(example_sentence) loaded_model.predict([example_sentence, self.empty_sentence]) loaded_model.predict([self.empty_sentence]) @pytest.mark.integration() def test_predict_with_prob(self, example_sentence, loaded_pretrained_model): loaded_pretrained_model.predict(example_sentence, return_probabilities_for_all_classes=True) assert len(example_sentence.get_labels(loaded_pretrained_model.label_type)) == len( loaded_pretrained_model.label_dictionary ) assert ( sum([label.score for label in example_sentence.get_labels(loaded_pretrained_model.label_type)]) > 1 - 1e-5 ) @pytest.mark.integration() def test_train_load_use_classifier_flair(self, results_base_path, corpus, example_sentence, train_test_sentence): flair.set_seed(123) embeddings = DocumentRNNEmbeddings([FlairEmbeddings("news-forward-fast")], 128, 1, False, 64, False, False) label_dict = corpus.make_label_dictionary(label_type=self.train_label_type) model = self.model_cls(embeddings=embeddings, label_dictionary=label_dict, label_type=self.train_label_type) trainer = ModelTrainer(model, corpus) trainer.train(results_base_path, max_epochs=2, shuffle=False) model.predict(train_test_sentence) for label in train_test_sentence.get_labels(self.train_label_type): assert label.value is not None assert 0.0 <= label.score <= 1.0 assert isinstance(label.score, float) del trainer, model, corpus loaded_model = self.model_cls.load(results_base_path / "final-model.pt") loaded_model.predict(example_sentence) loaded_model.predict([example_sentence, self.empty_sentence]) loaded_model.predict([self.empty_sentence])
4,076
39.77
118
py
flair
flair-master/tests/models/test_word_tagger.py
import pytest import flair from flair.embeddings import TransformerWordEmbeddings from flair.models import TokenClassifier from tests.model_test_utils import BaseModelTest class TestWordTagger(BaseModelTest): model_cls = TokenClassifier train_label_type = "pos" training_args = { "max_epochs": 2, "learning_rate": 0.1, "mini_batch_size": 2, } def has_embedding(self, sentence): for token in sentence: if token.get_embedding().cpu().numpy().size == 0: return False return None def build_model(self, embeddings, label_dict, **kwargs): model_args = dict(self.model_args) for k in kwargs: if k in model_args: del model_args[k] return self.model_cls( embeddings=embeddings, label_dictionary=label_dict, label_type=self.train_label_type, **model_args, **kwargs, ) @pytest.fixture() def corpus(self, tasks_base_path): return flair.datasets.UD_ENGLISH(tasks_base_path) @pytest.fixture() def embeddings(self): return TransformerWordEmbeddings("distilbert-base-uncased")
1,212
26.568182
67
py
flair
flair-master/tests/embeddings/test_word_embeddings.py
from typing import Any, Dict from flair.embeddings import MuseCrosslingualEmbeddings, NILCEmbeddings, WordEmbeddings from tests.embedding_test_utils import BaseEmbeddingsTest class TestWordEmbeddings(BaseEmbeddingsTest): embedding_cls = WordEmbeddings is_token_embedding = True is_document_embedding = False default_args = {"embeddings": "turian"} name_field = "embeddings" invalid_names = ["other", "not/existing/path/to/embeddings"] class TestMuseCrosslingualEmbeddings(BaseEmbeddingsTest): embedding_cls = MuseCrosslingualEmbeddings is_token_embedding = True is_document_embedding = False default_args: Dict[str, Any] = {} class TestNILCEmbeddings(BaseEmbeddingsTest): embedding_cls = NILCEmbeddings is_token_embedding = True is_document_embedding = False default_args = {"embeddings": "fasttext", "model": "cbow", "size": 50} valid_args = [{"embeddings": "glove"}] name_field = "embeddings" invalid_names = ["other", "not/existing/path/to/embeddings"]
1,033
30.333333
87
py
flair
flair-master/tests/embeddings/test_byte_pair_embeddings.py
from flair.embeddings import BytePairEmbeddings from tests.embedding_test_utils import BaseEmbeddingsTest class TestBytePairEmbeddings(BaseEmbeddingsTest): embedding_cls = BytePairEmbeddings is_token_embedding = True is_document_embedding = False default_args = {"language": "en"}
299
29
57
py
flair
flair-master/tests/embeddings/test_transformer_word_embeddings.py
import importlib.util import warnings import pytest import torch from PIL import Image from transformers.utils import is_detectron2_available from flair.data import BoundingBox, Dictionary, Sentence from flair.embeddings import TransformerJitWordEmbeddings, TransformerWordEmbeddings from flair.models import SequenceTagger from tests.embedding_test_utils import BaseEmbeddingsTest class TestTransformerWordEmbeddings(BaseEmbeddingsTest): embedding_cls = TransformerWordEmbeddings is_token_embedding = True is_document_embedding = False default_args = {"model": "distilbert-base-uncased", "allow_long_sentences": False} valid_args = [ {"layers": "-1,-2,-3,-4", "layer_mean": False}, {"layers": "all", "layer_mean": True}, {"layers": "all", "layer_mean": False}, {"layers": "all", "layer_mean": True, "subtoken_pooling": "mean"}, ] name_field = "embeddings" invalid_names = ["other", "not/existing/path/to/embeddings"] @pytest.mark.integration() def test_transformer_jit_embeddings(self, results_base_path): base_embeddings = TransformerWordEmbeddings( "distilbert-base-uncased", layers="-1,-2,-3,-4", layer_mean=False, allow_long_sentences=True ) sentence: Sentence = Sentence("I love Berlin, but Vienna is where my hearth is.") class JitWrapper(torch.nn.Module): def __init__(self, embedding: TransformerWordEmbeddings) -> None: super().__init__() self.embedding = embedding def forward( self, input_ids: torch.Tensor, token_lengths: torch.LongTensor, attention_mask: torch.Tensor, overflow_to_sample_mapping: torch.Tensor, word_ids: torch.Tensor, ): return self.embedding.forward( input_ids=input_ids, token_lengths=token_lengths, attention_mask=attention_mask, overflow_to_sample_mapping=overflow_to_sample_mapping, word_ids=word_ids, )["token_embeddings"] base_embeddings.embed(sentence) base_token_embedding = sentence[5].get_embedding().clone() sentence.clear_embeddings() tensors = base_embeddings.prepare_tensors([sentence]) # ensure that the prepared tensors is what we expect assert sorted(tensors.keys()) == [ "attention_mask", "input_ids", "overflow_to_sample_mapping", "token_lengths", "word_ids", ] wrapper = JitWrapper(base_embeddings) parameter_names, parameter_list = TransformerJitWordEmbeddings.parameter_to_list( base_embeddings, wrapper, [sentence] ) with warnings.catch_warnings(): warnings.simplefilter("ignore") script_module = torch.jit.trace(wrapper, parameter_list) jit_embeddings = TransformerJitWordEmbeddings.create_from_embedding( script_module, base_embeddings, parameter_names ) jit_embeddings.embed(sentence) jit_token_embedding = sentence[5].get_embedding().clone() assert torch.isclose(base_token_embedding, jit_token_embedding).all() sentence.clear_embeddings() # use a SequenceTagger to save and reload the embedding in the manner it is supposed to work example_tagger = SequenceTagger(embeddings=jit_embeddings, tag_dictionary=Dictionary(), tag_type="none") results_base_path.mkdir(exist_ok=True, parents=True) example_tagger.save(results_base_path / "tagger.pt") del example_tagger new_example_tagger = SequenceTagger.load(results_base_path / "tagger.pt") loaded_jit_embedding = new_example_tagger.embeddings loaded_jit_embedding.embed(sentence) loaded_jit_token_embedding = sentence[5].get_embedding().clone() sentence.clear_embeddings() assert torch.isclose(jit_token_embedding, loaded_jit_token_embedding).all() def test_transformers_context_expansion(self, results_base_path): emb = TransformerWordEmbeddings( "distilbert-base-uncased", use_context=True, use_context_separator=True, respect_document_boundaries=True ) # previous and next sentence as context sentence_previous = Sentence("How is it?") sentence_next = Sentence("Then again, maybe not...") # test expansion for sentence without context sentence = Sentence("This is great!") expanded, _ = emb._expand_sentence_with_context(sentence=sentence) assert " ".join([token.text for token in expanded]) == "[FLERT] This is great ! [FLERT]" # test expansion for with previous and next as context sentence = Sentence("This is great.") sentence._previous_sentence = sentence_previous sentence._next_sentence = sentence_next expanded, _ = emb._expand_sentence_with_context(sentence=sentence) assert ( " ".join([token.text for token in expanded]) == "How is it ? [FLERT] This is great . [FLERT] Then again , maybe not ..." ) # test expansion if first sentence is document boundary sentence = Sentence("This is great?") sentence_previous.is_document_boundary = True sentence._previous_sentence = sentence_previous sentence._next_sentence = sentence_next expanded, _ = emb._expand_sentence_with_context(sentence=sentence) assert ( " ".join([token.text for token in expanded]) == "[FLERT] This is great ? [FLERT] Then again , maybe not ..." ) # test expansion if we don't use context emb.context_length = 0 sentence = Sentence("I am here.") sentence._previous_sentence = sentence_previous sentence._next_sentence = sentence_next expanded, _ = emb._expand_sentence_with_context(sentence=sentence) assert " ".join([token.text for token in expanded]) == "I am here ." @pytest.mark.integration() def test_layoutlm_embeddings(self): sentence = Sentence(["I", "love", "Berlin"]) sentence[0].add_metadata("bbox", BoundingBox(0, 0, 10, 10)) sentence[1].add_metadata("bbox", (12, 0, 22, 10)) sentence[2].add_metadata("bbox", (0, 12, 10, 22)) emb = TransformerWordEmbeddings("microsoft/layoutlm-base-uncased", layers="-1,-2,-3,-4", layer_mean=True) emb.eval() emb.embed(sentence) @pytest.mark.integration() @pytest.mark.skipif( condition=not is_detectron2_available(), reason="layoutlmV2 requires detectron2 to be installed manually." ) def test_layoutlmv2_embeddings(self, tasks_base_path): with Image.open(tasks_base_path / "example_images" / "i_love_berlin.png") as img: img.load() img = img.convert("RGB") sentence = Sentence(["I", "love", "Berlin"]) sentence.add_metadata("image", img) sentence[0].add_metadata("bbox", BoundingBox(0, 0, 10, 10)) sentence[1].add_metadata("bbox", (12, 0, 22, 10)) sentence[2].add_metadata("bbox", (0, 12, 10, 22)) emb = TransformerWordEmbeddings("microsoft/layoutlmv2-base-uncased", layers="-1,-2,-3,-4", layer_mean=True) emb.eval() emb.embed(sentence) @pytest.mark.integration() def test_layoutlmv3_embeddings(self, tasks_base_path): with Image.open(tasks_base_path / "example_images" / "i_love_berlin.png") as img: img.load() img = img.convert("RGB") sentence = Sentence(["I", "love", "Berlin"]) sentence.add_metadata("image", img) sentence[0].add_metadata("bbox", BoundingBox(0, 0, 10, 10)) sentence[1].add_metadata("bbox", (12, 0, 22, 10)) sentence[2].add_metadata("bbox", (0, 12, 10, 22)) emb = TransformerWordEmbeddings("microsoft/layoutlmv3-base", layers="-1,-2,-3,-4", layer_mean=True) emb.eval() emb.embed(sentence) @pytest.mark.integration() def test_layoutlmv3_embeddings_with_long_context(self, tasks_base_path): with Image.open(tasks_base_path / "example_images" / "i_love_berlin.png") as img: img.load() img = img.convert("RGB") sentence = Sentence(["I", "love", "Berlin"] * 512) sentence.add_metadata("image", img) for i in range(512): sentence[i * 3].add_metadata("bbox", BoundingBox(0, 0, 10, 10)) sentence[i * 3 + 1].add_metadata("bbox", (12, 0, 22, 10)) sentence[i * 3 + 2].add_metadata("bbox", (0, 12, 0, 10)) emb = TransformerWordEmbeddings("microsoft/layoutlmv3-base", layers="-1,-2,-3,-4", layer_mean=True) emb.eval() emb.embed(sentence) @pytest.mark.integration() def test_ocr_embeddings_fails_when_no_bbox(self): sentence = Sentence(["I", "love", "Berlin"]) emb = TransformerWordEmbeddings("microsoft/layoutlm-base-uncased", layers="-1,-2,-3,-4", layer_mean=True) emb.eval() with pytest.raises(ValueError): emb.embed(sentence) @pytest.mark.integration() def test_layoutlm_embeddings_with_context_warns_user(self): sentence = Sentence(["I", "love", "Berlin"]) sentence[0].add_metadata("bbox", BoundingBox(0, 0, 10, 10)) sentence[1].add_metadata("bbox", (12, 0, 22, 10)) sentence[2].add_metadata("bbox", (0, 12, 10, 22)) with pytest.warns(UserWarning) as record: TransformerWordEmbeddings("microsoft/layoutlm-base-uncased", layers="-1,-2,-3,-4", use_context=True) assert len(record) == 1 assert "microsoft/layoutlm" in record[0].message.args[0] @pytest.mark.integration() def test_layoutlmv3_without_image_embeddings_fails(self): sentence = Sentence(["I", "love", "Berlin"]) sentence[0].add_metadata("bbox", BoundingBox(0, 0, 10, 10)) sentence[1].add_metadata("bbox", (12, 0, 22, 10)) sentence[2].add_metadata("bbox", (0, 12, 10, 22)) emb = TransformerWordEmbeddings("microsoft/layoutlmv3-base", layers="-1,-2,-3,-4", layer_mean=True) emb.eval() with pytest.raises(ValueError): emb.embed(sentence) @pytest.mark.skipif(importlib.util.find_spec("sacremoses") is None, reason="XLM-Embeddings require 'sacremoses'") def test_transformer_word_embeddings_forward_language_ids(self): cos = torch.nn.CosineSimilarity(dim=0, eps=1e-10) sent_en = Sentence(["This", "is", "a", "sentence"], language_code="en") sent_de = Sentence(["Das", "ist", "ein", "Satz"], language_code="de") embeddings = TransformerWordEmbeddings("xlm-mlm-ende-1024", layers="all", allow_long_sentences=False) embeddings.embed([sent_de, sent_en]) expected_similarities = [ 0.7102344036102295, 0.7598986625671387, 0.7437312602996826, 0.5584433674812317, ] for token_de, token_en, exp_sim in zip(sent_de, sent_en, expected_similarities): sim = cos(token_de.embedding, token_en.embedding).item() assert abs(exp_sim - sim) < 1e-5 def test_transformer_force_max_length(self): sentence: Sentence = Sentence("I love Berlin, but Vienna is where my hearth is.") short_embeddings = TransformerWordEmbeddings("distilbert-base-uncased", layers="-1,-2,-3,-4", layer_mean=False) long_embeddings = TransformerWordEmbeddings( "distilbert-base-uncased", layers="-1,-2,-3,-4", layer_mean=False, force_max_length=True ) short_tensors = short_embeddings.prepare_tensors([sentence]) long_tensors = long_embeddings.prepare_tensors([sentence]) for tensor in short_tensors.values(): if tensor.dim() > 1: # all tensors that have a sequence length need to be shorter assert tensor.shape[1] < 512 for tensor in long_tensors.values(): if tensor.dim() > 1: # all tensors that have a sequence length need to be exactly max length assert tensor.shape[1] == 512 short_embeddings.embed(sentence) short_embedding_0 = sentence[0].get_embedding() sentence.clear_embeddings() long_embeddings.embed(sentence) long_embedding_0 = sentence[0].get_embedding() # apparently the precision is not that high on cuda, hence the absolute tolerance needs to be higher. assert torch.isclose(short_embedding_0, long_embedding_0, atol=1e-4).all() def test_transformers_keep_tokenizer_when_saving(self, results_base_path): embeddings = TransformerWordEmbeddings("distilbert-base-uncased") results_base_path.mkdir(exist_ok=True, parents=True) initial_tagger_path = results_base_path / "initial_tokenizer.pk" reloaded_tagger_path = results_base_path / "reloaded_tokenizer.pk" initial_tagger = SequenceTagger(embeddings, Dictionary(), "ner") initial_tagger.save(initial_tagger_path) reloaded_tagger = SequenceTagger.load(initial_tagger_path) reloaded_tagger.save(reloaded_tagger_path) def test_transformers_keep_tokenizer_bloom_when_saving(self, results_base_path): embeddings = TransformerWordEmbeddings("Muennighoff/bloom-tiny-random") results_base_path.mkdir(exist_ok=True, parents=True) initial_tagger_path = results_base_path / "initial_tokenizer.pk" reloaded_tagger_path = results_base_path / "reloaded_tokenizer.pk" initial_tagger = SequenceTagger(embeddings, Dictionary(), "ner") initial_tagger.save(initial_tagger_path) reloaded_tagger = SequenceTagger.load(initial_tagger_path) reloaded_tagger.save(reloaded_tagger_path) def test_transformer_subword_token_mapping(self): sentence = Sentence("El pasto es verde.") embeddings = TransformerWordEmbeddings("PlanTL-GOB-ES/roberta-base-biomedical-es", layers="-1") embeddings.embed(sentence) @pytest.mark.skipif(importlib.util.find_spec("onnxruntime") is None, reason="Onnx export require 'onnxruntime'") def test_onnx_export_works(self, results_base_path): texts = [ "I live in Berlin", "I live in Vienna", "Berlin to Germany is like Vienna to Austria", ] normal_sentences = [Sentence(text) for text in texts] onnx_sentences = [Sentence(text) for text in texts] embeddings = TransformerWordEmbeddings("distilbert-base-uncased") results_base_path.mkdir(exist_ok=True, parents=True) onnx_embeddings = embeddings.export_onnx(results_base_path / "onnx-export.onnx", normal_sentences) embeddings.embed(normal_sentences) onnx_embeddings.embed(onnx_sentences) for sent_a, sent_b in zip(normal_sentences, onnx_sentences): for token_a, token_b in zip(sent_a, sent_b): assert torch.isclose(token_a.get_embedding(), token_b.get_embedding(), atol=1e-6).all()
15,110
45.352761
120
py
flair
flair-master/tests/embeddings/test_flair_embeddings.py
from flair.data import Dictionary, Sentence from flair.embeddings import ( DocumentLMEmbeddings, DocumentRNNEmbeddings, FlairEmbeddings, ) from flair.models import LanguageModel from tests.embedding_test_utils import BaseEmbeddingsTest class TestFlairEmbeddings(BaseEmbeddingsTest): embedding_cls = FlairEmbeddings is_token_embedding = True is_document_embedding = False default_args = {"model": "news-forward-fast"} name_field = "model" invalid_names = ["other", "not/existing/path/to/embeddings"] def test_fine_tunable_flair_embedding(self): language_model_forward = LanguageModel(Dictionary.load("chars"), is_forward_lm=True, hidden_size=32, nlayers=1) embeddings: DocumentRNNEmbeddings = DocumentRNNEmbeddings( [FlairEmbeddings(language_model_forward, fine_tune=True)], hidden_size=128, bidirectional=False, ) sentence: Sentence = Sentence("I love Berlin.") embeddings.embed(sentence) assert len(sentence.get_embedding()) == 128 assert len(sentence.get_embedding()) == embeddings.embedding_length sentence.clear_embeddings() assert len(sentence.get_embedding()) == 0 embeddings: DocumentLMEmbeddings = DocumentLMEmbeddings( [FlairEmbeddings(language_model_forward, fine_tune=True)] ) sentence: Sentence = Sentence("I love Berlin.") embeddings.embed(sentence) assert len(sentence.get_embedding()) == 32 assert len(sentence.get_embedding()) == embeddings.embedding_length sentence.clear_embeddings() assert len(sentence.get_embedding()) == 0 del embeddings
1,707
30.054545
119
py
flair
flair-master/tests/embeddings/test_simple_token_embeddings.py
from flair.data import Dictionary from flair.embeddings import CharacterEmbeddings, HashEmbeddings, OneHotEmbeddings from tests.embedding_test_utils import BaseEmbeddingsTest vocab_dictionary = Dictionary(add_unk=True) vocab_dictionary.add_item("I") vocab_dictionary.add_item("love") vocab_dictionary.add_item("berlin") class TestCharacterEmbeddings(BaseEmbeddingsTest): embedding_cls = CharacterEmbeddings is_token_embedding = True is_document_embedding = False default_args = {"path_to_char_dict": None} class TestOneHotEmbeddings(BaseEmbeddingsTest): embedding_cls = OneHotEmbeddings is_token_embedding = True is_document_embedding = False default_args = {"vocab_dictionary": vocab_dictionary} class TestHashEmbeddings(BaseEmbeddingsTest): embedding_cls = HashEmbeddings is_token_embedding = True is_document_embedding = False default_args = {"num_embeddings": 10}
923
29.8
82
py
flair
flair-master/tests/embeddings/test_stacked_embeddings.py
from flair.data import Sentence from flair.embeddings import ( FlairEmbeddings, StackedEmbeddings, TokenEmbeddings, WordEmbeddings, ) from flair.embeddings.base import load_embeddings def test_stacked_embeddings(): glove: TokenEmbeddings = WordEmbeddings("turian") flair_embedding: TokenEmbeddings = FlairEmbeddings("news-forward-fast") embeddings: StackedEmbeddings = StackedEmbeddings([glove, flair_embedding]) sentence: Sentence = Sentence("I love Berlin. Berlin is a great place to live.") embeddings.embed(sentence) for token in sentence.tokens: assert len(token.get_embedding()) == 1074 token.clear_embeddings() assert len(token.get_embedding()) == 0 del embeddings def test_stacked_embeddings_stay_the_same_after_saving_and_loading(): glove: TokenEmbeddings = WordEmbeddings("turian") flair_embedding: TokenEmbeddings = FlairEmbeddings("news-forward-fast") embeddings: StackedEmbeddings = StackedEmbeddings([glove, flair_embedding]) assert not embeddings.training sentence_old: Sentence = Sentence("I love Berlin") embeddings.embed(sentence_old) names_old = embeddings.get_names() embedding_length_old = embeddings.embedding_length save_data = embeddings.save_embeddings(use_state_dict=True) new_embeddings = load_embeddings(save_data) sentence_new: Sentence = Sentence("I love Berlin") new_embeddings.embed(sentence_new) names_new = new_embeddings.get_names() embedding_length_new = new_embeddings.embedding_length assert not new_embeddings.training assert names_old == names_new assert embedding_length_old == embedding_length_new for token_old, token_new in zip(sentence_old, sentence_new): assert (token_old.get_embedding(names_old) == token_new.get_embedding(names_new)).all()
1,851
33.296296
95
py
flair
flair-master/tests/embeddings/test_transformer_document_embeddings.py
from flair.data import Dictionary from flair.embeddings import TransformerDocumentEmbeddings from flair.models import TextClassifier from flair.nn import Classifier from tests.embedding_test_utils import BaseEmbeddingsTest class TestTransformerDocumentEmbeddings(BaseEmbeddingsTest): embedding_cls = TransformerDocumentEmbeddings is_document_embedding = True is_token_embedding = False default_args = {"model": "distilbert-base-uncased", "allow_long_sentences": False} valid_args = [ {"layers": "-1,-2,-3,-4", "layer_mean": False}, {"layers": "all", "layer_mean": True}, {"layers": "all", "layer_mean": False}, ] name_field = "embeddings" invalid_names = ["other", "not/existing/path/to/embeddings"] def test_if_loaded_embeddings_have_all_attributes(tasks_base_path): # dummy model with embeddings embeddings = TransformerDocumentEmbeddings( "distilbert-base-uncased", use_context=True, use_context_separator=False, ) model = TextClassifier(label_type="ner", label_dictionary=Dictionary(), embeddings=embeddings) # save the dummy and load it again model.save(tasks_base_path / "single.pt") loaded_single_task = Classifier.load(tasks_base_path / "single.pt") # check that context_length and use_context_separator is the same for both assert model.embeddings.context_length == loaded_single_task.embeddings.context_length assert model.embeddings.use_context_separator == loaded_single_task.embeddings.use_context_separator
1,549
37.75
104
py
flair
flair-master/tests/embeddings/__init__.py
0
0
0
py
flair
flair-master/tests/embeddings/test_tfidf_embeddings.py
from flair.data import Sentence from flair.embeddings import DocumentTFIDFEmbeddings from tests.embedding_test_utils import BaseEmbeddingsTest class TFIDFEmbeddingsTest(BaseEmbeddingsTest): embedding_cls = DocumentTFIDFEmbeddings is_document_embedding = True is_token_embedding = False default_args = { "train_dataset": [ Sentence("This is a sentence"), Sentence("This is another sentence"), Sentence("another a This I Berlin"), ] }
508
27.277778
57
py
flair
flair-master/tests/embeddings/test_document_transform_word_embeddings.py
from typing import Any, Dict, List from flair.embeddings import ( DocumentCNNEmbeddings, DocumentLMEmbeddings, DocumentPoolEmbeddings, DocumentRNNEmbeddings, FlairEmbeddings, TokenEmbeddings, WordEmbeddings, ) from tests.embedding_test_utils import BaseEmbeddingsTest word: TokenEmbeddings = WordEmbeddings("turian") flair_embedding: TokenEmbeddings = FlairEmbeddings("news-forward-fast") flair_embedding_back: TokenEmbeddings = FlairEmbeddings("news-backward-fast") class BaseDocumentsViaWordEmbeddingsTest(BaseEmbeddingsTest): is_document_embedding = True is_token_embedding = False base_embeddings: List[TokenEmbeddings] = [word, flair_embedding] def create_embedding_from_name(self, name: str): """Overwrite this method if it is more complex to load an embedding by name.""" assert self.name_field is not None kwargs = dict(self.default_args) kwargs.pop(self.name_field) return self.embedding_cls(name, **kwargs) # type: ignore[call-arg] def create_embedding_with_args(self, args: Dict[str, Any]): kwargs = dict(self.default_args) for k, v in args.items(): kwargs[k] = v return self.embedding_cls(self.base_embeddings, **kwargs) # type: ignore[call-arg] class TestDocumentLstmEmbeddings(BaseDocumentsViaWordEmbeddingsTest): embedding_cls = DocumentRNNEmbeddings default_args = { "hidden_size": 128, "bidirectional": False, } valid_args = [{"bidirectional": False}, {"bidirectional": True}] class TestDocumentPoolEmbeddings(BaseDocumentsViaWordEmbeddingsTest): embedding_cls = DocumentPoolEmbeddings default_args = { "fine_tune_mode": "nonlinear", } valid_args = [{"pooling": "mean"}, {"pooling": "max"}, {"pooling": "min"}] class TestDocumentCNNEmbeddings(BaseDocumentsViaWordEmbeddingsTest): embedding_cls = DocumentCNNEmbeddings default_args = { "kernels": ((50, 2), (50, 3)), } valid_args = [{"reproject_words_dimension": None}, {"reproject_words_dimension": 100}] class TestDocumentLMEmbeddings(BaseDocumentsViaWordEmbeddingsTest): embedding_cls = DocumentLMEmbeddings base_embeddings = [flair_embedding, flair_embedding_back] default_args: Dict[str, Any] = {}
2,301
33.358209
91
py
LSTM_Covariance
LSTM_Covariance-main/shallow_water/DA_preparation.py
# -*- coding: utf-8 -*- # assimilation shallow water import numpy as np def VAR_3D(xb,Y,H,B,R): #booleen=0 garde la trace dim_x = xb.size #dim_y = Y.size Y.shape = (Y.size,1) xb1=np.copy(xb) xb1.shape=(xb1.size,1) K=np.dot(B,np.dot(np.transpose(H),np.linalg.pinv(np.dot(H,np.dot(B,np.transpose(H)))+R))) #matrice de gain A=np.dot(np.dot((np.eye(dim_x)-np.dot(K,H)),B),np.transpose((np.eye(dim_x)-np.dot(K,H))))+np.dot(np.dot(K,R),np.transpose(K)) vect=np.dot(H,xb1) xa=np.copy(xb1+np.dot(K,(Y-vect))) return xa,A def x_to_y(X): # averaging in 2*2 windows (4 pixels) dim = int(X.shape[0]) dim = 20 Y = np.zeros((int(dim/2),int(dim/2))) for i in range(int(dim/2)): for j in range(int(dim/2)): Y[i,j] = X[2*i,2*j] + X[2*i+1,2*j] + X[2*i,2*j+1] + X[2*i+1,2*j+1] Y_noise = np.random.multivariate_normal(np.zeros(100),0.0000 * np.eye(100)) Y_noise.shape = (10,10) Y = Y + Y_noise return Y
1,041
25.05
129
py
LSTM_Covariance
LSTM_Covariance-main/shallow_water/prediction_plotting.py
# %% # -*- coding: utf-8 -*- """ Created on Wed Jan 6 13:39:58 2021 @author: siboc """ import numpy as np import scipy import math import matplotlib.pyplot as plt # data=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_1114.npy').astype(np.float32) data1=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_948.npy').astype(np.float32) train_data=np.concatenate((data,data1),axis=0) del data del data1 data2=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_947.npy').astype(np.float32) train_data=np.concatenate((train_data,data2),axis=0) del data2 data3=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_940.npy').astype(np.float32) train_data=np.concatenate((train_data,data3),axis=0) del data3 data4=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_941.npy').astype(np.float32) train_data=np.concatenate((train_data,data4),axis=0) del data4 data5=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_1111.npy').astype(np.float32) train_data=np.concatenate((train_data,data5),axis=0) del data5 data6=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_1112.npy').astype(np.float32) train_data=np.concatenate((train_data,data6),axis=0) del data6 data7=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_946.npy').astype(np.float32) train_data=np.concatenate((train_data,data7),axis=0) del data7 data8=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_945.npy').astype(np.float32) train_data=np.concatenate((train_data,data8),axis=0) del data8 data9=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_944.npy').astype(np.float32) train_data=np.concatenate((train_data,data9),axis=0) del data9 data10=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_943.npy').astype(np.float32) train_data=np.concatenate((train_data,data10),axis=0) del data10 data11=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_942.npy').astype(np.float32) train_data=np.concatenate((train_data,data11),axis=0) del data11 data12=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_1116.npy').astype(np.float32) train_data=np.concatenate((train_data,data12),axis=0) del data12 data13=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_1115.npy').astype(np.float32) train_data=np.concatenate((train_data,data13),axis=0) del data13 data14=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_1113.npy').astype(np.float32) train_data=np.concatenate((train_data,data14),axis=0) del data14 data15=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_939.npy').astype(np.float32) train_data=np.concatenate((train_data,data15),axis=0) del data15 data16=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_938.npy').astype(np.float32) train_data=np.concatenate((train_data,data16),axis=0) del data16 data17=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_937.npy').astype(np.float32) train_data=np.concatenate((train_data,data17),axis=0) del data17 data18=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_1117.npy').astype(np.float32) train_data=np.concatenate((train_data,data18),axis=0) del data18 data19=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_1118.npy').astype(np.float32) train_data=np.concatenate((train_data,data19),axis=0) del data19 # train_data=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_1179.npy').astype(np.float32) no # train_data=np.load('data2/trainset_withx_repeat_shwater3_uniform0011_test6_944.npy').astype(np.float32) # %% print("train_data shape: ",train_data.shape) # ############################################################################### obs = train_data.reshape((train_data.shape[0],200*1000+101)) del train_data # %% X=None y=None y=obs[:,-101:] y[:,-101:-1]=y[:,-101:-1]*1000 y[:,-1]=y[:,-1]/8 y=y.reshape((obs.shape[0],101)) X=obs[:,:200*1000].reshape((obs.shape[0],200,1000)) # X=std_scaler.transform(X) X=np.array([X[i].transpose() for i in range(X.shape[0])]) # input_data=X[:,:1000,:] input_data=X[:,:200,:] output_data=y ########################################################################## train_part = 0.97 threshold = int(train_part*obs.shape[0]) ########################################################################## # train_input = input_data[:threshold,:] # train_output = output_data[:threshold,:] # test_input = input_data [threshold:,:] # true_test_output = output_data[threshold:,:] # X1 = train_input # Y1 = train_output X2 = input_data true_test_output=output_data # train_input = input_data[:threshold,:] # train_output = output_data[:threshold,:] # test_input=input_data # true_test_output = output_data # X1 = train_input # Y1 = train_output # X2 = test_input ############################################################################ #def my_loss_fn(y_true, y_pred): # # return K.mean(K.abs(y_true - y_pred) * weight) # %% from tensorflow.keras.models import load_model # model1=load_model('data2/sequentiallstm200_b128_h200_norm_out_gen22.h5') model1=load_model('data2/sequentiallstm2222200_b128_h200_norm_out_gen22.h5') # Calculate predictions # PredValSet2 = model1.predict(X2.reshape(X2.shape[0],1000,200)) PredValSet2 = model1.predict(X2.reshape(X2.shape[0],200,200)) print("PredValSet2 shape: ",PredValSet2.shape) print("true_test_output shape: ",true_test_output.shape) # fig, ((ax1,ax2),(ax3,ax4),(ax5,ax6))=plt.subplots(nrows=3, ncols=2) # %% # sample1=np.random.randint(0,100,1) sample1=np.array([0]) print(sample1[0]) plt.plot(PredValSet2[:,sample1[0]],true_test_output[:,sample1[0]],'o', color='b',markersize=5) plt.plot(true_test_output[:,sample1[0]],true_test_output[:,sample1[0]],'-', color='r',linewidth=5) plt.xlabel('prediction',fontsize=22) plt.ylabel('true value',fontsize=22) # plt.xticks([-1.00,-0.50,0.00,0.50,1.00]) # plt.yticks([-1.00,-0.50,0.00,0.50,1.00]) plt.xticks(fontsize=22) plt.yticks(fontsize=22) # plt.legend() plt.show() # %% sample1=np.array([100]) print(sample1[0]) plt.plot(PredValSet2[:,sample1[0]],true_test_output[:,sample1[0]],'o', color='b',markersize=5) plt.plot(true_test_output[:,sample1[0]],true_test_output[:,sample1[0]],'-', color='r',linewidth=5) plt.xlabel('prediction',fontsize=22) plt.ylabel('true value',fontsize=22) # plt.xticks([-1.00,-0.50,0.00,0.50,1.00]) # plt.yticks([-1.00,-0.50,0.00,0.50,1.00]) plt.xticks(fontsize=22) plt.yticks(fontsize=22) plt.legend() plt.show() # %% # plt.plot(PredValSet2[:,sample1[0]],true_test_output[:,sample1[0]],'o', color='b',markersize=5) # plt.legend() # plt.show() # ax1.plot(PredValSet2[:,sample1[0]],true_test_output[:,sample1[0]],'o', color='b',markersize=5) # ax1.plot(true_test_output[:,sample1[0]],true_test_output[:,sample1[0]],'-', color='r',linewidth=5) # ax1.legend() # # sample2=np.random.randint(0,100,1) # sample2=np.array([82]) # while sample2[0]==sample1[0]: # sample2=np.random.randint(0,100,1) # print(sample2) # # plt.plot(PredValSet2[:,sample2[0]],true_test_output[:,sample2[0]],'o', color='b',markersize=5) # # plt.legend() # # plt.show() # # plt.xlim(-1, 1) # # plt.ylim(-1, 1) # # plt.plot(PredValSet1[:,1],true_test_output[:,1],'o', color='blue',markersize=5,label='lstm') # ax2.plot(PredValSet2[:,sample2[0]],true_test_output[:,sample2[0]],'o', color='b',markersize=5) # ax2.plot(true_test_output[:,sample2[0]],true_test_output[:,sample2[0]],'-', color='r',linewidth=5) # # plt.plot(PredValSet3[:,1],true_test_output[:,1],'o', color='green',markersize=5) # # plt.plot(PredValSet4[:,1],true_test_output[:,1],'o', color='c',markersize=5) # # plt.plot(PredValSet5[:,1],true_test_output[:,1],'o', color='m',markersize=5) # #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') # ax2.legend() # # sample3=np.random.randint(0,100,1) # sample3=np.array([84]) # while sample3[0]==sample2[0] or sample3[0]==sample1[0]: # sample3=np.random.randint(0,100,1) # print(sample3) # # # plt.xlim(-1, 1) # # # plt.ylim(-1, 1) # # # plt.plot(PredValSet1[:,2],true_test_output[:,2],'o', color='blue',markersize=5,label='lstm') # # plt.plot(PredValSet2[:,sample3[0]],true_test_output[:,sample3[0]],'o', color='b',markersize=5) # # # plt.plot(true_test_output[:,2],true_test_output[:,2],'o', color='r',markersize=5) # # # plt.plot(PredValSet3[:,2],true_test_output[:,2],'o', color='green',markersize=5) # # # plt.plot(PredValSet4[:,2],true_test_output[:,2],'o', color='c',markersize=5) # # # plt.plot(PredValSet5[:,2],true_test_output[:,2],'o', color='m',markersize=5) # # #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') # # plt.legend() # # plt.show() # # plt.plot(PredValSet1[:,2],true_test_output[:,2],'o', color='blue',markersize=5,label='lstm') # ax3.plot(PredValSet2[:,sample3[0]],true_test_output[:,sample3[0]],'o', color='b',markersize=5) # ax3.plot(true_test_output[:,sample3[0]],true_test_output[:,sample3[0]],'-', color='r',linewidth=5) # # plt.plot(PredValSet3[:,2],true_test_output[:,2],'o', color='green',markersize=5) # # plt.plot(PredValSet4[:,2],true_test_output[:,2],'o', color='c',markersize=5) # # plt.plot(PredValSet5[:,2],true_test_output[:,2],'o', color='m',markersize=5) # #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') # ax3.legend() # # sample4=np.random.randint(0,100,1) # sample4=np.array([86]) # while sample4[0]==sample2[0] or sample4[0]==sample1[0] or sample4[0]==sample3[0]: # sample4=np.random.randint(0,100,1) # print(sample4) # # # plt.xlim(-10, 10) # # # plt.ylim(-10, 10) # # # plt.plot(PredValSet1[:,3],true_test_output[:,3],'o',color='blue',markersize=5,label='lstm') # # plt.plot(PredValSet2[:,sample4[0]],true_test_output[:,sample4[0]],'o', color='b',markersize=5) # # # plt.plot(true_test_output[:,3],true_test_output[:,3],'o', color='r',markersize=5) # # # plt.plot(PredValSet3[:,3],true_test_output[:,3],'o', color='green',markersize=5) # # # plt.plot(PredValSet4[:,3],true_test_output[:,3],'o', color='c',markersize=5) # # # plt.plot(PredValSet5[:,3],true_test_output[:,3],'o', color='m',markersize=5) # # #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') # # plt.legend() # # plt.show() # # plt.xlim(-10, 10) # # plt.ylim(-10, 10) # # plt.plot(PredValSet1[:,3],true_test_output[:,3],'o',color='blue',markersize=5,label='lstm') # ax4.plot(PredValSet2[:,sample4[0]],true_test_output[:,sample4[0]],'o', color='b',markersize=5) # ax4.plot(true_test_output[:,sample4[0]],true_test_output[:,sample4[0]],'-', color='r',linewidth=5) # # plt.plot(PredValSet3[:,3],true_test_output[:,3],'o', color='green',markersize=5) # # plt.plot(PredValSet4[:,3],true_test_output[:,3],'o', color='c',markersize=5) # # plt.plot(PredValSet5[:,3],true_test_output[:,3],'o', color='m',markersize=5) # #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') # ax4.legend() # # plt.show() # # sample5=np.random.randint(0,100,1) # sample5=np.array([88]) # while sample5[0]==sample2[0] or sample5[0]==sample1[0] or sample5[0]==sample3[0] or sample5[0]==sample4[0]: # sample5=np.random.randint(0,100,1) # print(sample5) # ax5.plot(PredValSet2[:,sample5[0]],true_test_output[:,sample5[0]],'o', color='b',markersize=5) # ax5.plot(true_test_output[:,sample5[0]],true_test_output[:,sample5[0]],'-', color='r',linewidth=5) # ax5.legend() # sample6=np.array([99]) # print(sample6) # ax6.plot(PredValSet2[:,sample6[0]],true_test_output[:,sample6[0]],'o', color='b',markersize=5) # ax6.plot(true_test_output[:,sample6[0]],true_test_output[:,sample6[0]],'-', color='r',linewidth=5) # # plt.plot(PredValSet3[:,3],true_test_output[:,3],'o', color='green',markersize=5) # # plt.plot(PredValSet4[:,3],true_test_output[:,3],'o', color='c',markersize=5) # # plt.plot(PredValSet5[:,3],true_test_output[:,3],'o', color='m',markersize=5) # #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') # ax6.legend() # fig.supxlabel('Prediction') # fig.supylabel('true value') # plt.tight_layout() # plt.show() # # plt.plot(true_test_output[:,0],color='b',label='r0') # # plt.plot(true_test_output[:,1],color='r',label='r1') # # plt.plot(true_test_output[:,2],color='y',label='r2') # # plt.legend() # # plt.show() # ##########################################################################################" # # predint = model.predict(train_input[:3000]) # # trueint = train_output[:3000] # # plt.plot(predint[:,3],trueint[:,3],'o', color='blue',markersize=5) # # #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') # # plt.show()
12,529
31.973684
109
py
LSTM_Covariance
LSTM_Covariance-main/shallow_water/data_generation_no_v_20.py
#shallow water propagation """ Solution of Shallow-water equations using a Python class. Adapted for Python training course at CNRS from https://github.com/mrocklin/ShallowWater/ Dmitry Khvorostyanov, 2015 CNRS/LMD/IPSL, dmitry.khvorostyanov @ lmd.polytechnique.fr """ import time from pylab import * import matplotlib.gridspec as gridspec from shallowwater import * # import imageio from constructB import * from DA_preparation import * import os if not os.path.isdir('uniform01_2'): os.makedirs('uniform01_2') print("30/05/2021") class shallow_dynamique(object): time = 0 plt = [] fig = [] def __init__(self, x=[],y=[],u=zeros((100,100)),v = zeros((100,100)),h=ones((100,100)),dx=0.01,dt=0.0001, N=100,L=1., g=1., b=2.0): # How define no default argument before? self.g = g self.b = b self.L=L self.N=N # limits for h,u,v self.dx=dx self.dt=dt self.x,self.y = mgrid[:self.N,:self.N] self.u=u self.v=v self.h=h # self.h= ones((self.N,self.N)) #self.lims = [(self.h_ini-self.Hp,self.h_ini+self.Hp),(-0.02,0.02),(-0.02,0.02)] def dxy(self, A, axis=0): """ Compute derivative of array A using balanced finite differences Axis specifies direction of spatial derivative (d/dx or d/dy) dA[i]/dx = (A[i+1] - A[i-1] ) / 2dx """ return (roll(A, -1, axis) - roll(A, 1, axis)) / (self.dx*2.) # roll: shift the array axis=0 shift the horizontal axis def d_dx(self, A): return self.dxy(A,1) def d_dy(self, A): return self.dxy(A,0) def d_dt(self, h, u, v): """ http://en.wikipedia.org/wiki/Shallow_water_equations#Non-conservative_form """ for x in [h, u, v]: # type check assert isinstance(x, ndarray) and not isinstance(x, matrix) g,b,dx = self.g, self.b, self.dx du_dt = -g*self.d_dx(h) - b*u dv_dt = -g*self.d_dy(h) - b*v H = 0 #h.mean() - our definition of h includes this term dh_dt = -self.d_dx(u * (h)) - self.d_dy(v * (h)) return dh_dt, du_dt, dv_dt def evolve(self): """ Evolve state (h, u, v) forward in time using simple Euler method x_{N+1} = x_{N} + dx/dt * d_t """ dh_dt, du_dt, dv_dt = self.d_dt(self.h, self.u, self.v) dt = self.dt self.h += dh_dt * dt self.u += du_dt * dt self.v += dv_dt * dt self.time += dt return self.h, self.u, self.v ############################################################################ #define H (from x to y in the one dimensional space) H_1D_uv = np.zeros((100,400)) for i in range(10): for j in range(10): H_1D_uv[i*10+j,2*i*20+2*j] = 1 H_1D_uv[i*10+j,(2*i+1)*20+2*j] = 1 H_1D_uv[i*10+j,2*i*20+2*j+1] = 1 H_1D_uv[i*10+j,(2*i+1)*20+2*j+1] = 1 H_1strow = np.concatenate((H_1D_uv,np.zeros((100,400))),axis = 1) H_2ndrow = np.concatenate((np.zeros((100,400)),H_1D_uv),axis = 1) H = np.concatenate((H_1strow,H_2ndrow),axis = 0) ############################################################################ iteration_times=1000 parameter_size=100+1 trainning_set = np.zeros((1,iteration_times*200+parameter_size)) index=210+60*26 try: for ii in range(0,30000): if ii%100==0: print(ii) ############################################################################# # define the R matrix # D = 0.001*np.random.lognormal(0, 0.7, 100) D = 0.001*np.random.uniform(0.01, 0.1, 100) r = np.random.uniform(1, 8) # same observation error covariance for u and v (D is the lower unit triangle, Balgovind is the diagonal triangle, R is the SPD or HPD) R = np.dot(np.dot(np.sqrt(np.diag(D)),Balgovind(10,r)[:100,:100]),np.sqrt(np.diag(D))) import copy D_copy=D.copy() D_copy.shape=(10,10) parameters=np.zeros((1,parameter_size)) parameters[0,:100]=D parameters[0,100:]=r # v = np.random.uniform(0,100.) # R=v*R ############################################################################## SW = shallow(u=0.001*np.identity(20),v=0.001*np.identity(20),px=10,py=10,N = 20,R=10) ############################################################################################### h_ini = SW.h # SW = shallow_dynamique(u = np.zeros((SW.h.shape[0],SW.h.shape[1])), # v = np.zeros((SW.h.shape[0],SW.h.shape[1])), # h = h_ini) y_obs = np.zeros((H.shape[0],iteration_times)) ############################################################################################### for i in range(iteration_times): print("sample time: ",ii) print("iteration time: ",i) print(f"index {index}") SW.evolve() Y = np.dot(H, np.concatenate((SW.u.ravel(),SW.v.ravel())). reshape(2*SW.u.size,1))#generate observations #R = np.dot(np.dot(np.sqrt(np.diag(D)),np.eye(100)),np.sqrt(np.diag(D))) Y[:100] += np.random.multivariate_normal(np.zeros(100),R).reshape(100,1)# u: update Y[100:] += np.random.multivariate_normal(np.zeros(100),R).reshape(100,1)# v: update # Y: 200*1 #y_obs: 200*iteration_times # ValueError: could not broadcast input array from shape (200,1) into shape (200): # https://stackoverflow.com/a/39825046/10349608 y_obs[:,[i]]=Y train_row=y_obs.ravel() train_row.shape = (1,train_row.size) train_row=np.concatenate((train_row,parameters),axis=1) trainning_set=np.concatenate((trainning_set,train_row),axis=0) if (ii+1)%500==0: index=index+1 trainning_set = trainning_set[1:,:] np.save(f"uniform01_2/trainset_withx_repeat_shwater3_uniform0011_{index}.npy", trainning_set) trainning_set = np.zeros((1,iteration_times*200+parameter_size)) # if (ii+1)%1000==0: # np.savetxt("uniform01_2/trainset_withx_repeat_shwater3_uniform0011_total_test6.csv", trainning_set, delimiter=",") except Exception as e: np.savetxt("uniform01_2/trainset_withx_repeat_shwater3_uniform0011_total_test6_1.csv", trainning_set, delimiter=",") except KeyboardInterrupt: np.savetxt("uniform01_2/trainset_withx_repeat_shwater3_uniform0011_total_test6_1.csv", trainning_set, delimiter=",") except UserAbort: np.savetxt("uniform01_2/trainset_withx_repeat_shwater3_uniform0011_total_test6_1.csv", trainning_set, delimiter=",")
7,057
25.335821
176
py
LSTM_Covariance
LSTM_Covariance-main/shallow_water/shallowwater.py
""" Solution of Shallow-water equations using a Python class. Adapted for Python training course at CNRS from https://github.com/mrocklin/ShallowWater/ Dmitry Khvorostyanov, 2015 CNRS/LMD/IPSL, dmitry.khvorostyanov @ lmd.polytechnique.fr """ import time from pylab import * import matplotlib.gridspec as gridspec import numpy as np #construct background states, observations with error def x_to_y(X): # averaging in 2*2 windows (4 pixels) dim = X.shape[0] dim = 20 Y = np.zeros((dim/2,dim/2)) for i in range(dim/2): for j in range(dim/2): Y[i,j] = X[2*i,2*j] + X[2*i+1,2*j] + X[2*i,2*j+1] + X[2*i+1,2*j+1] Y_noise = np.random.multivariate_normal(np.zeros(100),0.0000 * np.eye(100)) Y_noise.shape = (10,10) Y = Y + Y_noise return Y class shallow(object): # domain #N = 100 #L = 1. #dx = L / N #dt = dx / 100. # Initial Conditions #u = zeros((N,N)) # velocity in x direction #v = zeros((N,N)) # velocity in y direction #h_ini = 1. #h = h_ini * ones((N,N)) # pressure deviation (like height) #x,y = mgrid[:N,:N] time = 0 plt = [] fig = [] def __init__(self, x=[],y=[],h_ini = 1.,u=[],v = [],dx=0.01,dt=0.0001, N=100,L=1., px=50, py=50, R=100, Hp=0.1, g=1., b=2.): # How define no default argument before? # add a perturbation in pressure surface self.px, self.py = px, py self.R = R self.Hp = Hp # Physical parameters self.g = g self.b = b self.L=L self.N=N # limits for h,u,v #self.dx = self.L / self.N # a changer #self.dt = self.dx / 100. self.dx=dx self.dt=dt self.x,self.y = mgrid[:self.N,:self.N] self.u=zeros((self.N,self.N)) self.v=zeros((self.N,self.N)) self.h_ini=h_ini self.h=self.h_ini * ones((self.N,self.N)) rr = (self.x-px)**2 + (self.y-py)**2 self.h[rr<R] = self.h_ini + Hp #set initial conditions self.lims = [(self.h_ini-self.Hp,self.h_ini+self.Hp),(-0.02,0.02),(-0.02,0.02)] def dxy(self, A, axis=0): """ Compute derivative of array A using balanced finite differences Axis specifies direction of spatial derivative (d/dx or d/dy) dA[i]/dx = (A[i+1] - A[i-1] ) / 2dx """ return (roll(A, -1, axis) - roll(A, 1, axis)) / (self.dx*2.) # roll: shift the array axis=0 shift the horizontal axis def d_dx(self, A): return self.dxy(A,1) def d_dy(self, A): return self.dxy(A,0) def d_dt(self, h, u, v): """ http://en.wikipedia.org/wiki/Shallow_water_equations#Non-conservative_form """ for x in [h, u, v]: # type check assert isinstance(x, ndarray) and not isinstance(x, matrix) g,b,dx = self.g, self.b, self.dx du_dt = -g*self.d_dx(h) - b*u dv_dt = -g*self.d_dy(h) - b*v H = 0 #h.mean() - our definition of h includes this term dh_dt = -self.d_dx(u * (H+h)) - self.d_dy(v * (H+h)) return dh_dt, du_dt, dv_dt def evolve(self): """ Evolve state (h, u, v) forward in time using simple Euler method x_{N+1} = x_{N} + dx/dt * d_t """ dh_dt, du_dt, dv_dt = self.d_dt(self.h, self.u, self.v) dt = self.dt self.h += dh_dt * dt self.u += du_dt * dt self.v += dv_dt * dt self.time += dt return self.h, self.u, self.v ## def plot(self,tit=None,autolims=False): ## """Plot u,v,h at current state.""" ## ## self.fig.append(figure()) ## for i,v in enumerate([self.h,self.u,self.v]): ## if autolims: ## vmin,vmax = None, None ## else: ## vmin, vmax = self.lims[i][0], self.lims[i][1] ## ## self.fig[-1].add_subplot(3,1,i+1) ## ## self.plt.append(pcolormesh(v, vmin=vmin, vmax=vmax)) ## colorbar(shrink=0.9) ## ## if i==0: ## if tit is None: ## self.tit = title('At time %f'% self.time) ## else: ## self.tit = title(tit) ## ## ## def animate(self): ## """Plot u,v,h at current state.""" ## ## for i,v in enumerate([self.h,self.u,self.v]): ## self.plt[i].set_array(v.ravel()) ## ## if i==0: ## self.tit.set_text('At time %f'%self.time) if __name__ == '__main__': #run the current script iteration_times= 500 SW = shallow(N=20,px=10,py=10,R=10.) # chose a point (x,y) to check the evolution x=10 y=10 #SW.plot() # u_vect=np.zeros(iteration_times) # v_vect=np.zeros(iteration_times) # h_vect=np.zeros(iteration_times) # for i in range(iteration_times): # SW.evolve() # u_vect[i]=SW.u[x][y] # v_vect[i]=SW.v[x][y] # h_vect[i]=SW.h[x][y] # #SW.animate() # # if i % 100 == 0: # print ('time %f'%SW.time) ## SW.fig[-1].savefig('sw_%.3d.png'% i) ## ## show() # print SW.time ## plt.subplot(311) ## plt.imshow( SW.h) ## set_title("Title for first plot") ## ## ## plt.subplot(312) ## plt.imshow( SW.u) ## ## ## ## plt.subplot(313) ## plt.imshow( SW.v) ## ## plt.show() gs = gridspec.GridSpec(2, 2, width_ratios=[1, 1], height_ratios=[1, 1] ) fig = plt.figure() #fig = plt.figure() t=SW.time fig = plt.figure() t=SW.time fig.suptitle('At time: T=%1.3f'%t, fontsize=16) ax1 = plt.subplot(gs[0]) ax1.set_title("h") im=ax1.imshow( SW.h) plt.colorbar(im) ax2 = plt.subplot(gs[1]) ax2.set_title("u") im=ax2.imshow( SW.u) plt.colorbar(im) ax3 = plt.subplot(gs[2]) ax3.set_title("v") im=ax3.imshow( SW.v) plt.colorbar(im) plt.show() gs = gridspec.GridSpec(1, 2, width_ratios=[1, 1], ) fig = plt.figure() #fig = plt.figure() t=SW.time ################################################################################ #u_t=SW.u[[45,46,47,48,49,51,52,53,54,55], :][:, [45,46,47,48,49,51,52,53,54,55]] #v_t=SW.v[[45,46,47,48,49,51,52,53,54,55], :][:, [45,46,47,48,49,51,52,53,54,55]] iteration_times = 5000 for i in range(iteration_times): SW.evolve() if i%100 ==0: u_t = SW.u v_t = SW.v # fig = plt.figure() # t=SW.time # fig.suptitle('At time: T=%1.3f'%t, fontsize=16) # ax1 = plt.subplot(gs[0]) # ax1.set_title("u") # im=ax1.imshow( u_t,interpolation='none') # plt.colorbar(im) # # ax2 = plt.subplot(gs[1]) # ax2.set_title("v") # im=ax2.imshow( v_t,interpolation='none') # plt.colorbar(im) # # plt.show() im = plt.imshow(SW.h, interpolation = "None") plt.colorbar(im) plt.title("$h (t = 0.05)$") #plt.savefig("Figures/h_t005.eps", format = "eps") plt.show() plt.close() im = plt.imshow(SW.u, interpolation = "None") plt.colorbar(im) plt.title("$u (t = 0.05)$") #plt.savefig("Figures/u_t005.eps", format = "eps") plt.show() plt.close() im = plt.imshow(SW.v, interpolation = "None") plt.colorbar(im) plt.title("$v (t = 0.05)$") #plt.savefig("Figures/v_t005.eps", format = "eps") plt.show() plt.close() # i = 300 # xbu = np.loadtxt("data/SW_b_u_"+str(i)+".txt") # xbv = np.loadtxt("data/SW_b_v_"+str(i)+".txt") # xbh = np.loadtxt("data/SW_b_h_"+str(i)+".txt") # # im = plt.imshow(xbu, interpolation = "None") # plt.colorbar(im) # plt.title("$u_b (t = 0.05)$") # #plt.savefig("Figures/u_b005.eps", format = "eps") # plt.show() # plt.close() # # im = plt.imshow(xbv, interpolation = "None") # plt.colorbar(im) # plt.title("$v_b (t = 0.05)$") # #plt.savefig("Figures/v_b005.eps", format = "eps") # plt.show() # plt.close() # # im = plt.imshow(xbh, interpolation = "None") # plt.colorbar(im) # plt.title("$h_b (t = 0.05)$") # #plt.savefig("Figures/h_b005.eps", format = "eps") # plt.show() # plt.close() # # yu = x_to_y(SW.u) # yv = x_to_y(SW.v) # # im = plt.imshow(yu, interpolation = "None") # plt.colorbar(im) # plt.title("$y_{t,u} (t = 0.05)$") # #plt.savefig("Figures/yt_u005.eps", format = "eps") # plt.show() # plt.close() # # im = plt.imshow(yv, interpolation = "None") # plt.colorbar(im) # plt.title("$y_{t,v} (t = 0.05)$") # #plt.savefig("Figures/yt_v005.eps", format = "eps") # plt.show() # plt.close() # # ybu = np.loadtxt("data/SW_yu_"+str(i)+".txt") # ybv = np.loadtxt("data/SW_yv_"+str(i)+".txt") # # im = plt.imshow(ybu, interpolation = "None") # plt.colorbar(im) # plt.title("$y_{b,u} (t = 0.05)$") # #plt.savefig("Figures/yb_u005.eps", format = "eps") # plt.show() # plt.close() # # im = plt.imshow(ybv, interpolation = "None") # plt.colorbar(im) # plt.title("$y_{b,v} (t = 0.05)$") # #plt.savefig("Figures/yb_v005.eps", format = "eps") # plt.show() # plt.close()
9,745
24.989333
169
py
LSTM_Covariance
LSTM_Covariance-main/shallow_water/simulated_data_generation.py
#shallow water propagation """ Solution of Shallow-water equations using a Python class. Adapted for Python training course at CNRS from https://github.com/mrocklin/ShallowWater/ Dmitry Khvorostyanov, 2015 CNRS/LMD/IPSL, dmitry.khvorostyanov @ lmd.polytechnique.fr """ import time from pylab import * import matplotlib.gridspec as gridspec from shallowwater import * # import imageio from constructB import * from DA_preparation import * import os if not os.path.isdir('uniform01_2'): os.makedirs('uniform01_2') print("30/05/2021") class shallow_dynamique(object): time = 0 plt = [] fig = [] def __init__(self, x=[],y=[],u=zeros((100,100)),v = zeros((100,100)),h=ones((100,100)),dx=0.01,dt=0.0001, N=100,L=1., g=1., b=2.0): # How define no default argument before? self.g = g self.b = b self.L=L self.N=N # limits for h,u,v self.dx=dx self.dt=dt self.x,self.y = mgrid[:self.N,:self.N] self.u=u self.v=v self.h=h # self.h= ones((self.N,self.N)) #self.lims = [(self.h_ini-self.Hp,self.h_ini+self.Hp),(-0.02,0.02),(-0.02,0.02)] def dxy(self, A, axis=0): """ Compute derivative of array A using balanced finite differences Axis specifies direction of spatial derivative (d/dx or d/dy) dA[i]/dx = (A[i+1] - A[i-1] ) / 2dx """ return (roll(A, -1, axis) - roll(A, 1, axis)) / (self.dx*2.) # roll: shift the array axis=0 shift the horizontal axis def d_dx(self, A): return self.dxy(A,1) def d_dy(self, A): return self.dxy(A,0) def d_dt(self, h, u, v): """ http://en.wikipedia.org/wiki/Shallow_water_equations#Non-conservative_form """ for x in [h, u, v]: # type check assert isinstance(x, ndarray) and not isinstance(x, matrix) g,b,dx = self.g, self.b, self.dx du_dt = -g*self.d_dx(h) - b*u dv_dt = -g*self.d_dy(h) - b*v H = 0 #h.mean() - our definition of h includes this term dh_dt = -self.d_dx(u * (h)) - self.d_dy(v * (h)) return dh_dt, du_dt, dv_dt def evolve(self): """ Evolve state (h, u, v) forward in time using simple Euler method x_{N+1} = x_{N} + dx/dt * d_t """ dh_dt, du_dt, dv_dt = self.d_dt(self.h, self.u, self.v) dt = self.dt self.h += dh_dt * dt self.u += du_dt * dt self.v += dv_dt * dt self.time += dt return self.h, self.u, self.v ############################################################################ #define H (from x to y in the one dimensional space) H_1D_uv = np.zeros((100,400)) for i in range(10): for j in range(10): H_1D_uv[i*10+j,2*i*20+2*j] = 1 H_1D_uv[i*10+j,(2*i+1)*20+2*j] = 1 H_1D_uv[i*10+j,2*i*20+2*j+1] = 1 H_1D_uv[i*10+j,(2*i+1)*20+2*j+1] = 1 H_1strow = np.concatenate((H_1D_uv,np.zeros((100,400))),axis = 1) H_2ndrow = np.concatenate((np.zeros((100,400)),H_1D_uv),axis = 1) H = np.concatenate((H_1strow,H_2ndrow),axis = 0) ############################################################################ iteration_times=1000 parameter_size=100+1 trainning_set = np.zeros((1,iteration_times*200+parameter_size)) index=0 try: for ii in range(0,30000): if ii%100==0: print(ii) ############################################################################# # define the R matrix # D = 0.001*np.random.lognormal(0, 0.7, 100) D = 0.001*np.random.uniform(0.01, 0.1, 100) r = np.random.uniform(1, 8) # same observation error covariance for u and v (D is the lower unit triangle, Balgovind is the diagonal triangle, R is the SPD or HPD) R = np.dot(np.dot(np.sqrt(np.diag(D)),Balgovind(10,r)[:100,:100]),np.sqrt(np.diag(D))) import copy D_copy=D.copy() D_copy.shape=(10,10) parameters=np.zeros((1,parameter_size)) parameters[0,:100]=D parameters[0,100:]=r # v = np.random.uniform(0,100.) # R=v*R ############################################################################## SW = shallow(u=np.zeros((20,20)),v=np.zeros((20,20)),px=10,py=10,N = 20,R=10) ############################################################################################### h_ini = SW.h # SW = shallow_dynamique(u = np.zeros((SW.h.shape[0],SW.h.shape[1])), # v = np.zeros((SW.h.shape[0],SW.h.shape[1])), # h = h_ini) y_obs = np.zeros((H.shape[0],iteration_times)) ############################################################################################### for i in range(iteration_times): print("sample time: ",ii) print("iteration time: ",i) print(f"index {index}") SW.evolve() Y = np.dot(H, np.concatenate((SW.u.ravel(),SW.v.ravel())). reshape(2*SW.u.size,1))#generate observations #R = np.dot(np.dot(np.sqrt(np.diag(D)),np.eye(100)),np.sqrt(np.diag(D))) Y[:100] += np.random.multivariate_normal(np.zeros(100),R).reshape(100,1)# u: update Y[100:] += np.random.multivariate_normal(np.zeros(100),R).reshape(100,1)# v: update # Y: 200*1 #y_obs: 200*iteration_times # ValueError: could not broadcast input array from shape (200,1) into shape (200): # https://stackoverflow.com/a/39825046/10349608 y_obs[:,[i]]=Y train_row=y_obs.ravel() train_row.shape = (1,train_row.size) train_row=np.concatenate((train_row,parameters),axis=1) trainning_set=np.concatenate((trainning_set,train_row),axis=0) if (ii+1)%500==0: index=index+1 trainning_set = trainning_set[1:,:] np.save(f"uniform01_2/trainset_withx_repeat_shwater3_uniform0011_{index}.npy", trainning_set) trainning_set = np.zeros((1,iteration_times*200+parameter_size)) # if (ii+1)%1000==0: # np.savetxt("uniform01_2/trainset_withx_repeat_shwater3_uniform0011_total_test6.csv", trainning_set, delimiter=",") except Exception as e: np.savetxt("uniform01_2/trainset_withx_repeat_shwater3_uniform0011_total_test6_1.csv", trainning_set, delimiter=",") except KeyboardInterrupt: np.savetxt("uniform01_2/trainset_withx_repeat_shwater3_uniform0011_total_test6_1.csv", trainning_set, delimiter=",") except UserAbort: np.savetxt("uniform01_2/trainset_withx_repeat_shwater3_uniform0011_total_test6_1.csv", trainning_set, delimiter=",")
7,041
25.276119
176
py
LSTM_Covariance
LSTM_Covariance-main/shallow_water/constructB.py
# coding: utf8 #construction of matrix B and special H with measure on the boarder import numpy as np import math from scipy.linalg import sqrtm from shallowwater import * ##def B_Balgovind(n,Sigma,L): ## Gamma = np.identity(n) ## for i in xrange(n): ## for j in xrange(n): ## Gamma[i,j] = ( 1. + abs(i-j)/L)*np.exp(-abs(i-j)/L) ## B = np.dot(Sigma,np.dot(Gamma,Sigma)) ## return B def get_index_2d (dim,n): #get caratesian coordinate j=n % dim j=j/1. #float( i) i=(n-j)/dim return (i,j)# pourquoi float? #identite def identiity(n):#n : taille de vecteur xb B=np.eye(n) return B #Blgovind def Balgovind(dim,L): sub_B=np.zeros((dim**2,dim**2)) for i in range(dim**2): (a1,b1)=get_index_2d(dim,i) for j in range(dim**2): (a2,b2)=get_index_2d(dim,j) #reprends les donnees caracterisennes r=math.sqrt((a1-a2)**2+(b1-b2)**2) sub_B[i,j]=(1+r/L)*(math.exp(-r/L)) B1=np.concatenate((sub_B, np.zeros((dim**2,dim**2))), axis=1) B2=np.concatenate(( np.zeros((dim**2,dim**2)), sub_B),axis=1) B=np.concatenate((B1,B2), axis=0)# a changer construction matrice B return B def Gaussian(dim,L): sub_B=np.zeros((dim**2,dim**2)) for i in range(dim**2): (a1,b1)=get_index_2d(dim,i) for j in range(dim**2): (a2,b2)=get_index_2d(dim,j) #reprends les donnees caracterisennes r=math.sqrt((a1-a2)**2+(b1-b2)**2) sub_B[i,j]=math.exp(-r**2/(2*L**2)) B1=np.concatenate((sub_B, np.zeros((dim**2,dim**2))), axis=1) B2=np.concatenate(( np.zeros((dim**2,dim**2)), sub_B),axis=1) B=np.concatenate((B1,B2), axis=0)# a changer construction matrice B return B def expontielle(dim,L): sub_B=np.zeros((dim**2,dim**2)) for i in range(dim**2): (a1,b1)=get_index_2d(dim,i) for j in range(dim**2): (a2,b2)=get_index_2d(dim,j) #reprends les donnees caracterisennes r=math.sqrt((a1-a2)**2+(b1-b2)**2) sub_B[i,j]=math.exp(-r/L) B1=np.concatenate((sub_B, np.zeros((dim**2,dim**2))), axis=1) B2=np.concatenate(( np.zeros((dim**2,dim**2)), sub_B),axis=1) B=np.concatenate((B1,B2), axis=0)# a changer construction matrice B return B def bord_M_aleatoire(dimension, proba): M=np.zeros((dimension**2,dimension**2)) for i in range(dimension**2): for j in range(dimension**2): if j % 10==0 or j % 10==9: M[i,j]=np.random.binomial(1, proba) elif j<=9 or j>=90: M[i,j]=np.random.binomial(1, proba) return M def cov_to_cor(B): inv_diag_B=np.linalg.inv(sqrtm(np.diag(np.diag(B)))) inv_diag_B=np.copy(inv_diag_B.real) cor_B=np.dot(inv_diag_B,np.dot(B,inv_diag_B)) return cor_B ######################################################################## #covariance 1d def Balgovind_1D(dim,L): B=np.zeros((dim,dim)) for i in range(dim): for j in range(dim): r=abs(i-j)*1. B[i,j]=(1+r/L)*(math.exp(-r/L)) return B def expontielle_1D(dim,L): B=np.zeros((dim,dim)) for i in range(dim): for j in range(dim): r=abs(i-j)*1. B[i,j]=math.exp(-r/L) return B def Gaussian_1D(dim,L): B=np.zeros((dim,dim)) for i in range(dim): for j in range(dim): r=abs(i-j)*1. B[i,j]=math.exp(-r**2/(2*L**2)) return B #Blgovind def Balgovind_noniso(dim,L,rayon): #diag_vect: vector of dim, variance of each point sub_B=np.zeros((dim**2,dim**2)) for i in range(dim**2): (a1,b1)=get_index_2d(dim,i) for j in range(dim**2): (a2,b2)=get_index_2d(dim,j) #reprends les donnees caracterisennes r=math.sqrt((a1-a2)**2+(b1-b2)**2) sub_B[i,j]=(1+r/L)*(math.exp(-r/L)) rr1 = math.sqrt((a1-5)**2 + (b1-5)**2) #cercle in the middle rr2 = math.sqrt((a2-5)**2 + (b2-5)**2) if rr1<=rayon: #inside the cercle sub_B[i,j]=sub_B[i,j]*4 else: sub_B[i,j]=sub_B[i,j]*0.25 if rr2<=rayon: sub_B[i,j]=sub_B[i,j]*4 else: sub_B[i,j]=sub_B[i,j]*0.25 B1=np.concatenate((sub_B, np.zeros((dim**2,dim**2))), axis=1) B2=np.concatenate(( np.zeros((dim**2,dim**2)), sub_B),axis=1) B=np.concatenate((B1,B2), axis=0)# a changer construction matrice B return B if __name__ == "__main__": im = plt.imshow(0.01*Balgovind(10,3)[:100,:100]) #im = plt.imshow(0.01*Balgovind(10,3)) plt.colorbar(im) #plt.savefig("Figures/R_bal_3.eps", format ='eps')
4,880
31.324503
84
py
LSTM_Covariance
LSTM_Covariance-main/shallow_water/shallowwater_lstm1000_model.py
# -*- coding: utf-8 -*- """ Created on Wed Jan 6 13:39:58 2021 @author: siboc """ import numpy as np import scipy import math from tensorflow.keras.models import Sequential from tensorflow.keras.layers import Dense from sklearn.metrics import r2_score import tensorflow as tf import tensorflow.keras.backend as K # check scikit-learn version # check scikit-learn version import sklearn from sklearn.linear_model import LinearRegression from sklearn.datasets import make_regression from sklearn.neighbors import KNeighborsRegressor from sklearn.tree import DecisionTreeRegressor import pandas as pd # import keras from tensorflow.keras.models import Sequential,load_model from tensorflow.keras.layers import LSTM,Dropout from tensorflow.keras.layers import TimeDistributed from tensorflow.keras.callbacks import EarlyStopping from tensorflow.keras.callbacks import ModelCheckpoint from tensorflow.keras.optimizers import Adam import os import json import pickle # gpus = tf.config.experimental.list_physical_devices('GPU') # if gpus: # try: # # Currently, memory growth needs to be the same across GPUs # for gpu in gpus: # tf.config.experimental.set_memory_growth(gpu, True) # logical_gpus = tf.config.experimental.list_logical_devices('GPU') # print(len(gpus), "Physical GPUs,", len(logical_gpus), "Logical GPUs") # except RuntimeError as e: # # Memory growth must be set before GPUs have been initialized # print(e) physical_devices = tf.config.list_physical_devices('GPU') try: tf.config.experimental.set_memory_growth(physical_devices[0:3], True) except: # Invalid device or cannot modify virtual devices once initialized. pass #======================================================================= # Generator class DataGenerator(tf.keras.utils.Sequence): # Generates data for keras # list_IDs: all IDs/all files # list_IDs_temp: studying batch IDs def __init__(self, list_IDs,batch_size=1,dim=(1000,200),n_channels=1, shuffle=True): self.dim=dim self.batch_size=batch_size # self.labels=labels self.list_IDs=list_IDs self.n_channels=n_channels # self.n_classes=n_classes self.shuffle=shuffle self.on_epoch_end() def __len__(self): return int(np.floor(len(self.list_IDs)/self.batch_size)) def __getitem__(self, index): print(index) # we do not know the whole length of the matrix in a file without loading all of the data ?????????? indexes=self.indexes[index*self.batch_size:(index+1)*self.batch_size] list_IDs_temp = [self.list_IDs[k] for k in indexes] X, y = self.__data_generation(list_IDs_temp) return X,y def on_epoch_end(self): 'Updates indexes after each epoch' self.indexes = np.arange(len(self.list_IDs)) if self.shuffle == True: np.random.shuffle(self.indexes) def __data_generation(self, list_IDs_temp): for i,ID in enumerate(list_IDs_temp): # open the file obs = np.load('uniform020/trainset_withx_repeat_shwater3_uniform0011_test6_'+str(ID)+'.npy') obs_size=obs.shape[0] print("ID: ",ID) print("obs_size: ",obs_size) X=np.empty((obs_size,200*1000)) y=np.empty((obs_size,101)) y=obs[:,-101:] y[:,-101:-1]=y[:,-101:-1]*1000 y[:,-1]=y[:,-1]/8 X=obs[:,:200*1000] X=X.reshape((obs_size,200,1000)) X=np.array([X[j].transpose() for j in range(obs_size)]) return X,y print("31") print("============================================================") print("============================================================") print("============================================================") print("============================================================") print("============================================================") print("============================================================") print("============================================================") # #====================== Read file list_IDs ================================================= partition=np.array([i for i in range(211)]) # #====================== Parameters ================================================= Params={'dim':(1000,200), 'batch_size':1, 'n_channels':1, 'shuffle':True} # train_data1000 = np.array(pd.read_csv('0001D07/trainset_withx_repeat_shwater3_0001D07_total_3.csv',delimiter=",", # header=None, # index_col=False)) # obs = train_data1000.reshape((train_data1000.shape[0],200*1000+101)) # X=np.empty((train_data1000.shape[0],200*1000)) # y=np.empty((train_data1000.shape[0],101)) # y=obs[:,-101:] # X=obs[:,:200*1000].reshape((train_data1000.shape[0],200,1000)) # X=np.array([X[i].transpose() for i in range(X.shape[0])]) # input_data=X[:,:,:200] # output_data=y # #====================== Generators ================================================= train_part = 0.97 threshold = int(train_part*len(partition)) # input_generator=DataGenerator(partition,**Params) training_generator=DataGenerator(partition[:threshold],**Params) validation_generator=DataGenerator(partition[threshold:],**Params) #============================= Model Design ========================================== if not os.path.isdir('data2'): os.makedirs('data2') # try: hidden_size=200 model = Sequential() model.add(LSTM(hidden_size,input_shape=(1000,200))) model.add(Dense(101)) # model=load_model('data2/sequentiallstm1000_b128_h200_norm_out_new_model.h5',compile = False) model.compile(loss='mean_squared_error', optimizer='Adam', metrics=['mae']) print(model.summary()) es=EarlyStopping(monitor='val_loss',mode='min',verbose=1,patience=100) # modelcheckpoint mc=ModelCheckpoint('data2/sequentiallstm1000_b128_h200_norm_out_new_model.h5',monitor='val_loss',mode='min',save_best_only=True,verbose=1) # history=model.fit(input_data, output_data, validation_split=train_part, epochs=100, batch_size=128, verbose=1,callbacks=[es,mc]) history=model.fit(x=training_generator, validation_data=validation_generator, epochs=1000, validation_batch_size=5,verbose=1, callbacks=[es,mc]) # history=model.fit(x=training_generator, validation_data=validation_generator, epochs=1000, use_multiprocessing=True, workers=6,verbose=1,callbacks=[es,mc]) # model.save('save_data/sequentiallstm2') model.save('data2/sequentiallstm1000_b128_h200_norm_out_new_model_f.h5') # https://stackoverflow.com/a/44674337/10349608 with open('data2/sequentiallstm1000_b128_h200_norm_out_new_model_history.pickle', 'wb') as file_his: pickle.dump(history.history, file_his)
6,880
30.135747
157
py
LSTM_Covariance
LSTM_Covariance-main/shallow_water/shallowwater_lstm200_model.py
# -*- coding: utf-8 -*- """ Created on Wed Jan 6 13:39:58 2021 @author: siboc """ import numpy as np import scipy import math from tensorflow.keras.models import Sequential from tensorflow.keras.layers import Dense from sklearn.metrics import r2_score import tensorflow as tf import tensorflow.keras.backend as K # check scikit-learn version # check scikit-learn version import sklearn from sklearn.linear_model import LinearRegression from sklearn.datasets import make_regression from sklearn.neighbors import KNeighborsRegressor from sklearn.tree import DecisionTreeRegressor import pandas as pd # import keras from tensorflow.keras.models import Sequential,load_model from tensorflow.keras.layers import LSTM,Dropout from tensorflow.keras.layers import TimeDistributed from tensorflow.keras.callbacks import EarlyStopping from tensorflow.keras.callbacks import ModelCheckpoint from tensorflow.keras.optimizers import Adam import os import json import pickle # gpus = tf.config.experimental.list_physical_devices('GPU') # if gpus: # try: # # Currently, memory growth needs to be the same across GPUs # for gpu in gpus: # tf.config.experimental.set_memory_growth(gpu, True) # logical_gpus = tf.config.experimental.list_logical_devices('GPU') # print(len(gpus), "Physical GPUs,", len(logical_gpus), "Logical GPUs") # except RuntimeError as e: # # Memory growth must be set before GPUs have been initialized # print(e) physical_devices = tf.config.list_physical_devices('GPU') try: tf.config.experimental.set_memory_growth(physical_devices[0:3], True) except: # Invalid device or cannot modify virtual devices once initialized. pass #======================================================================= # Generator class DataGenerator(tf.keras.utils.Sequence): # Generates data for keras # list_IDs: all IDs/all files # list_IDs_temp: studying batch IDs def __init__(self, list_IDs,batch_size=1,dim=(1000,200),n_channels=1, shuffle=True): self.dim=dim self.batch_size=batch_size # self.labels=labels self.list_IDs=list_IDs self.n_channels=n_channels # self.n_classes=n_classes self.shuffle=shuffle self.on_epoch_end() def __len__(self): return int(np.floor(len(self.list_IDs)/self.batch_size)) def __getitem__(self, index): print(index) # we do not know the whole length of the matrix in a file without loading all of the data ?????????? indexes=self.indexes[index*self.batch_size:(index+1)*self.batch_size] list_IDs_temp = [self.list_IDs[k] for k in indexes] X, y = self.__data_generation(list_IDs_temp) return X,y def on_epoch_end(self): 'Updates indexes after each epoch' self.indexes = np.arange(len(self.list_IDs)) if self.shuffle == True: np.random.shuffle(self.indexes) def __data_generation(self, list_IDs_temp): for i,ID in enumerate(list_IDs_temp): # open the file obs = np.load('uniform020/trainset_withx_repeat_shwater3_uniform0011_test6_'+str(ID)+'.npy') obs_size=obs.shape[0] print("ID: ",ID) print("obs_size: ",obs_size) X=np.empty((obs_size,200*1000)) y=np.empty((obs_size,101)) y=obs[:,-101:] y[:,-101:-1]=y[:,-101:-1]*1000 y[:,-1]=y[:,-1]/8 X=obs[:,:200*1000] X=X.reshape((obs_size,200,1000))[:,:,:200] X=np.array([X[j].transpose() for j in range(obs_size)]) return X,y print("31") print("============================================================") print("============================================================") print("============================================================") print("============================================================") print("============================================================") print("============================================================") print("============================================================") # #====================== Read file list_IDs ================================================= partition=np.array([i for i in range(211)]) # #====================== Parameters ================================================= Params={'dim':(1000,200), 'batch_size':1, 'n_channels':1, 'shuffle':True} # train_data1000 = np.array(pd.read_csv('0001D07/trainset_withx_repeat_shwater3_0001D07_total_3.csv',delimiter=",", # header=None, # index_col=False)) # obs = train_data1000.reshape((train_data1000.shape[0],200*1000+101)) # X=np.empty((train_data1000.shape[0],200*1000)) # y=np.empty((train_data1000.shape[0],101)) # y=obs[:,-101:] # X=obs[:,:200*1000].reshape((train_data1000.shape[0],200,1000)) # X=np.array([X[i].transpose() for i in range(X.shape[0])]) # input_data=X[:,:,:200] # output_data=y # #====================== Generators ================================================= train_part = 0.97 threshold = int(train_part*len(partition)) # input_generator=DataGenerator(partition,**Params) training_generator=DataGenerator(partition[:threshold],**Params) validation_generator=DataGenerator(partition[threshold:],**Params) #============================= Model Design ========================================== if not os.path.isdir('data2'): os.makedirs('data2') # try: hidden_size=200 model = Sequential() model.add(LSTM(hidden_size,input_shape=(200,200))) model.add(Dense(101)) # model=load_model('data2/sequentiallstm2222200_b128_h200_norm_out_gen22.h5',compile = False) model.compile(loss='mean_squared_error', optimizer='Adam', metrics=['mae']) print(model.summary()) es=EarlyStopping(monitor='val_loss',mode='min',verbose=1,patience=100) # modelcheckpoint mc=ModelCheckpoint('data2/sequentiallstm2222200_b128_h200_norm_out_gen22.h5',monitor='val_loss',mode='min',save_best_only=True,verbose=1) # history=model.fit(input_data, output_data, validation_split=train_part, epochs=100, batch_size=128, verbose=1,callbacks=[es,mc]) history=model.fit(x=training_generator, validation_data=validation_generator, epochs=1000, validation_batch_size=5,verbose=1, callbacks=[es,mc]) # history=model.fit(x=training_generator, validation_data=validation_generator, epochs=1000, use_multiprocessing=True, workers=6,verbose=1,callbacks=[es,mc]) # model.save('save_data/sequentiallstm2') model.save('data2/sequentiallstm2222200_b128_h200_norm_out_gen22_f.h5') # https://stackoverflow.com/a/44674337/10349608 with open('data2/sequentiallstm2222200_b128_h200_norm_out_gen22_history.pickle', 'wb') as file_his: pickle.dump(history.history, file_his)
6,889
29.622222
157
py
LSTM_Covariance
LSTM_Covariance-main/lorenz/lstmR_d05R_plotting.py
# -*- coding: utf-8 -*- """ Created on Wed Jan 6 13:39:58 2021 @author: siboc """ import numpy as np import matplotlib.pyplot as plt # check scikit-learn version # check scikit-learn version import pandas as pd # def data_set_order(file): # train_data = np.array(pd.read_csv(file)) # r0=train_data[:,:1001] # r1=train_data[:,1001:2002] # r2=train_data[:,2002:3003] # r3=train_data[:,3003:]/10 # train_data=np.insert(r0,[i+1 for i in range(r0.shape[1])],r1,axis=1) # train_data=np.insert(train_data,[(i+1)*2 for i in range(int(train_data.shape[1]/2))],r2,axis=1) # train_data=np.concatenate((train_data,r3),axis=1) # return train_data def data_set_order(file): train_data = np.array(pd.read_csv(file))[:-2,:] r0=train_data[:,:1001] r1=train_data[:,1001:2002] r2=train_data[:,2002:3003] r3=train_data[:,3003:] r3[:,-1]=r3[:,-1]/100 train_data=np.insert(r0,[i+1 for i in range(r0.shape[1])],r1,axis=1) train_data=np.insert(train_data,[(i+1)*2 for i in range(int(train_data.shape[1]/2))],r2,axis=1) train_data=np.concatenate((train_data,r3),axis=1) return train_data # def data_set_order(file): # train_data = np.array(pd.read_csv(file))[:-2,:] # r0=train_data[:,:1001][:,:201] # r1=train_data[:,1001:2002][:,:201] # r2=train_data[:,2002:3003][:,:201] # r3=train_data[:,3003:] # r3[:,-1]=r3[:,-1]/100 # train_data=np.insert(r0,[i+1 for i in range(r0.shape[1])],r1,axis=1) # train_data=np.insert(train_data,[(i+1)*2 for i in range(int(train_data.shape[1]/2))],r2,axis=1) # train_data=np.concatenate((train_data,r3),axis=1) # return train_data train_data = data_set_order('lorenz_cov_train_v2/trainset_withx_steps1000_11.csv')[:10000,:] print("train_data shape: ",train_data.shape) # ############################################################################### # LSTM1000 input_data = train_data[:,0:1001*3] output_data = train_data[:,1001*3:] # LSTM200 # input_data = train_data[:,0:603] # output_data = train_data[:,603:] ######################################################################## train_part = 0.97 # threshold = int(train_part*train_data.shape[0]) threshold=10000 ########################################################################## test_input = input_data[:threshold,:] true_test_output = output_data[:threshold,:] # test_input = input_data [threshold:,:] # true_test_output = output_data[threshold:,:] # X1 = train_input # Y1 = train_output X2 = test_input print("X2 shape: ",X2.shape[0]) #Y2 = ValidationSet_Y ############################################################################ R=np.load('label_data/di05_original_version_R_all_10000.npy') PredValSet2=np.zeros((R.shape[0],4)) for i in range(R.shape[0]): print(i) r0=R[i,0,1] r1=R[i,0,2] r2=R[i,1,2] r3=np.trace(R[i,:,:])/3 r=np.array([[r0,r1,r2,r3]]) if i==0: PredValSet2=r.copy() else: PredValSet2=np.concatenate((PredValSet2,r),axis = 0) #def my_loss_fn(y_true, y_pred): # # return K.mean(K.abs(y_true - y_pred) * weight) # from tensorflow.keras.models import load_model # model1=load_model('data2/sequentiallstm1000_ing_f.h5') # model1=load_model('data2/sequentiallstm200_ing_f.h5') # Calculate predictions # PredValSet2 = model1.predict(X2.reshape(X2.shape[0],1001,3)) # PredValSet2 = model1.predict(X2.reshape(X2.shape[0],201,3)) # PredTestSet = model.predict(X1) # PredValSet = model.predict(X2) # Save predictions #np.savetxt("numerique/trainresults_raindebit.csv", PredTestSet, delimiter=",") #np.savetxt("numerique/valresults_raindebit.csv", PredValSet, delimiter=",") # plt.plot(history.history['loss']) # plt.plot(history.history['val_loss']) # plt.title('Model loss') # plt.ylabel('Loss') # plt.xlabel('Epoch') # plt.legend(['Train', 'Test'], loc='upper left') # #plt.savefig('figure_dp/loss_trace.eps', format='eps',bbox_inches='tight') # plt.show() #plt.plot(true_test_output[:,1],'r',label = "true") #plt.plot(PredValSet[:,1],label = "model") #plt.title("1st coeff linear") #plt.legend() #plt.show() #deep_error = [] # #for i in range(150): # # deep_error.append(np.linalg.norm(PredValSet[:,i]-true_test_output[:,i])) # #print('deep_error',deep_error) # %% # plt.xlim(-1, 1) # plt.ylim(-1, 1) # plt.plot(PredValSet1[:,0],true_test_output[:,0],'o', color='blue',markersize=5,label='lstm') # %% plt.plot(PredValSet2[:,0]/PredValSet2[:,3],true_test_output[:,0],'o', color='b',markersize=5) plt.plot(true_test_output[:,0],true_test_output[:,0],'-', color='r',linewidth=5) plt.xlabel('prediction',fontsize=22) plt.ylabel('true value',fontsize=22) plt.xticks([-2.00,-1.00,0.00,1.00,2.00]) plt.yticks([-1.00,-0.50,0.00,0.50,1.00]) plt.xticks(fontsize=22) plt.yticks(fontsize=22) plt.legend() plt.show() # %% # plt.xlim(-1, 1) # plt.ylim(-1, 1) # plt.plot(PredValSet1[:,0],true_test_output[:,0],'o', color='blue',markersize=5,label='lstm') # plt.plot(PredValSet2[:,0],true_test_output[:,0],'o', color='b',markersize=5) # plt.plot(true_test_output[:,0],true_test_output[:,0],'o', color='r',markersize=5) # plt.plot(PredValSet3[:,0],true_test_output[:,0],'o', color='green',markersize=5) # plt.plot(PredValSet4[:,0],true_test_output[:,0],'o', color='c',markersize=5) # plt.plot(PredValSet5[:,0],true_test_output[:,0],'o', color='m',markersize=5) #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') # plt.legend() # plt.show() # plt.xlim(-1, 1) # plt.ylim(-1, 1) # plt.plot(PredValSet1[:,1],true_test_output[:,1],'o', color='blue',markersize=5,label='lstm') # %% plt.plot(PredValSet2[:,1]/PredValSet2[:,3],true_test_output[:,1],'o', color='b',markersize=5) plt.plot(true_test_output[:,1],true_test_output[:,1],'-', color='r',linewidth=5) # plt.xlabel('prediction',fontsize=22) # plt.ylabel('true value',fontsize=22) # plt.xticks([-1.00,-0.50,0.00,0.50,1.00]) # plt.yticks([-1.00,-0.50,0.00,0.50,1.00]) # plt.xticks(fontsize=22) # plt.yticks(fontsize=22) # plt.plot(PredValSet3[:,1],true_test_output[:,1],'o', color='green',markersize=5) # plt.plot(PredValSet4[:,1],true_test_output[:,1],'o', color='c',markersize=5) # plt.plot(PredValSet5[:,1],true_test_output[:,1],'o', color='m',markersize=5) #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') plt.legend() plt.show() # %% # plt.xlim(-1, 1) # plt.ylim(-1, 1) # plt.plot(PredValSet1[:,1],true_test_output[:,1],'o', color='blue',markersize=5,label='lstm') # plt.plot(PredValSet2[:,1],true_test_output[:,1],'o', color='b',markersize=5) # plt.plot(true_test_output[:,1],true_test_output[:,1],'o', color='r',markersize=5) # plt.plot(PredValSet3[:,1],true_test_output[:,1],'o', color='green',markersize=5) # plt.plot(PredValSet4[:,1],true_test_output[:,1],'o', color='c',markersize=5) # plt.plot(PredValSet5[:,1],true_test_output[:,1],'o', color='m',markersize=5) #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') # plt.legend() # plt.show() # plt.xlim(-1, 1) # plt.ylim(-1, 1) # plt.plot(PredValSet1[:,2],true_test_output[:,2],'o', color='blue',markersize=5,label='lstm') # plt.plot(PredValSet2[:,2],true_test_output[:,2],'o', color='b',markersize=5) # plt.plot(true_test_output[:,2],true_test_output[:,2],'o', color='r',markersize=5) # plt.plot(PredValSet3[:,2],true_test_output[:,2],'o', color='green',markersize=5) # plt.plot(PredValSet4[:,2],true_test_output[:,2],'o', color='c',markersize=5) # plt.plot(PredValSet5[:,2],true_test_output[:,2],'o', color='m',markersize=5) #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') # plt.legend() # plt.show() # plt.plot(PredValSet1[:,2],true_test_output[:,2],'o', color='blue',markersize=5,label='lstm') # %% plt.plot(PredValSet2[:,2]/PredValSet2[:,3],true_test_output[:,2],'o', color='b',markersize=5) plt.plot(true_test_output[:,2],true_test_output[:,2],'-', color='r',linewidth=5) plt.xlabel('prediction',fontsize=22) plt.ylabel('true value',fontsize=22) plt.xticks([-1.00,-0.50,0.00,0.50,1.00]) plt.yticks([-1.00,-0.50,0.00,0.50,1.00]) plt.xticks(fontsize=22) plt.yticks(fontsize=22) plt.legend() plt.show() # %% # plt.xlim(-10, 10) # plt.ylim(-10, 10) # plt.plot(PredValSet1[:,3],true_test_output[:,3],'o',color='blue',markersize=5,label='lstm') # plt.plot(PredValSet2[:,3],true_test_output[:,3],'o', color='b',markersize=5) # plt.plot(true_test_output[:,3],true_test_output[:,3],'o', color='r',markersize=5) # plt.plot(PredValSet3[:,3],true_test_output[:,3],'o', color='green',markersize=5) # plt.plot(PredValSet4[:,3],true_test_output[:,3],'o', color='c',markersize=5) # plt.plot(PredValSet5[:,3],true_test_output[:,3],'o', color='m',markersize=5) # plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') # plt.legend() # plt.show() # plt.xlim(-10, 10) # plt.ylim(-10, 10) # plt.plot(PredValSet1[:,3],true_test_output[:,3],'o',color='blue',markersize=5,label='lstm') # %% plt.plot(PredValSet2[:,3],true_test_output[:,3]*100,'o', color='b',markersize=5) plt.plot(true_test_output[:,3]*100,true_test_output[:,3]*100,'-', color='r',linewidth=5) # plt.xlabel('prediction',fontsize=22) # plt.ylabel('true value',fontsize=22) # # plt.xticks([-1.00,-0.50,0.00,0.50,1.00]) # plt.xticks(fontsize=22) # plt.yticks(fontsize=22) plt.legend() plt.show() # %% # plt.plot(true_test_output[:,0],color='b',label='r0') # plt.plot(true_test_output[:,1],color='r',label='r1') # plt.plot(true_test_output[:,2],color='y',label='r2') # plt.legend() # plt.show() ##########################################################################################" # predint = model.predict(train_input[:3000]) # trueint = train_output[:3000] # plt.plot(predint[:,3],trueint[:,3],'o', color='blue',markersize=5) # #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') # plt.show()
9,710
29.731013
101
py
LSTM_Covariance
LSTM_Covariance-main/lorenz/lorenz_lstm1000.py
# -*- coding: utf-8 -*- """ Created on Wed Jan 6 13:39:58 2021 @author: siboc """ import numpy as np import scipy import math import matplotlib.pyplot as plt from keras.models import Sequential from keras.layers import Dense from sklearn.metrics import r2_score import tensorflow as tf import keras.backend as K import sys # check scikit-learn version # check scikit-learn version import sklearn from sklearn.linear_model import LinearRegression from sklearn.datasets import make_regression from sklearn.neighbors import KNeighborsRegressor from sklearn.tree import DecisionTreeRegressor import pandas as pd def data_set_order(file): train_data = np.array(pd.read_csv(file))[:-2,:] r0=train_data[:,:1001] r1=train_data[:,1001:2002] r2=train_data[:,2002:3003] r3=train_data[:,3003:] r3[:,-1]=r3[:,-1]/100 train_data=np.insert(r0,[i+1 for i in range(r0.shape[1])],r1,axis=1) train_data=np.insert(train_data,[(i+1)*2 for i in range(int(train_data.shape[1]/2))],r2,axis=1) train_data=np.concatenate((train_data,r3),axis=1) return train_data #input train_data = data_set_order('lorenz_cov_train_v2/trainset_withx_steps1000_11.csv') print("train_data shape: ",train_data.shape) print(f"training dataset size: {train_data.shape[0]*0.9}") print(f"validation dataset size: {train_data.shape[0]*0.1}") sys.exit() # train_data1 = data_set_order('lorenz_cov_train/trainset_withx_repeat10bis3.csv') # print("train_data1 shape: ",train_data1.shape) # train_data2 = data_set_order('lorenz_cov_train/trainset_withx_repeat10bis4.csv') # print("train_data2 shape: ",train_data2.shape) # train_data3 = data_set_order('lorenz_cov_train/trainset_withx_repeat10bis5.csv') # print("train_data3 shape: ",train_data3.shape) # train_data4 = data_set_order('lorenz_cov_train/trainset_withx_repeat10bis6.csv') # print("train_data4 shape: ",train_data4.shape) # train_data5 = data_set_order('lorenz_cov_train/trainset_withx_repeat10bis7.csv') # print("train_data5 shape: ",train_data5.shape) # train_data6 = data_set_order('lorenz_cov_train/trainset_withx_repeat10bis8.csv') # print("train_data6 shape: ",train_data6.shape) #size: num_steps*3,r1,r2,r3,v ######################################################################################### #train_data = np.array(pd.read_csv('data_1000steps/trainset_withx_1000steps.csv')) # # #train_data1 = np.array(pd.read_csv('data_1000steps/trainset_withx_1000stepsbis1.csv')) # #train_data2 = np.array(pd.read_csv('data_1000steps/trainset_withx_1000stepsbis2.csv')) # #train_data3 = np.array(pd.read_csv('data_1000steps/trainset_withx_1000stepsbis3.csv')) # train_data = np.concatenate((train_data6,train_data5),axis = 0) # train_data = np.concatenate((train_data,train_data4),axis = 0) # train_data = np.concatenate((train_data,train_data3),axis = 0) # train_data = np.concatenate((train_data,train_data2),axis = 0) # train_data = np.concatenate((train_data,train_data1),axis = 0) # train_data = np.concatenate((train_data,train_data0),axis = 0) # train_data=train_data[:120000,:] #weightstrain_data[:,604:] np.random.shuffle(train_data ) input_data = train_data[:,0:3003] output_data = train_data[:,3003:] ######################################################################## train_part = 0.97 threshold = int(train_part*train_data.shape[0]) ########################################################################## train_input = input_data[:threshold] print("input_data shape: ",input_data.shape) train_output = output_data[:threshold] print("output_data shape: ",output_data.shape) test_input = input_data [threshold:] true_test_output = output_data[threshold:] X1 = train_input Y1 = train_output X2 = test_input #Y2 = ValidationSet_Y ############################################################################ #def my_loss_fn(y_true, y_pred): # # return K.mean(K.abs(y_true - y_pred) * weight) # ======================================================================================== from keras.layers import LSTM,Dropout from keras.layers import TimeDistributed from keras.callbacks import EarlyStopping from keras.callbacks import ModelCheckpoint from keras.optimizers import Adam # save data import os import json import pickle if not os.path.isdir('save_data_v2'): os.makedirs('save_data_v2') hidden_size=200 input_sample=input_data.shape[0] #for one sample output_sample=output_data.shape[0] input_data=input_data.reshape(input_sample,1001,3) #201 is the time steps in data_generation output_data=output_data.reshape(output_sample,4) use_dropout=True model = Sequential() model.add(LSTM(hidden_size,input_shape=(1001,3))) model.add(Dense(4)) # opt = Adam(lr=0.0001) model.compile(loss='mean_squared_error', optimizer='Adam', metrics=['mae']) print(model.summary()) es=EarlyStopping(monitor='val_loss',mode='min',verbose=1,patience=50) # modelcheckpoint mc=ModelCheckpoint('save_data_v2/sequentiallstm1000_ing.h5',monitor='val_loss',mode='min',save_best_only=True,verbose=1) history=model.fit(input_data, output_data, validation_split=0.1, epochs=100, batch_size=128, verbose=1,callbacks=[es,mc]) # model.save('save_data/sequentiallstm2') model.save('save_data_v2/sequentiallstm1000_ing_f.h5') # https://stackoverflow.com/a/44674337/10349608 with open('save_data_v2/sequentiallstm1000_ing_history.pickle', 'wb') as file_his: pickle.dump(history.history, file_his) # Calculate predictions PredTestSet = model.predict(X1.reshape(X1.shape[0],1001,3)) PredValSet = model.predict(X2.reshape(X2.shape[0],1001,3)) plt.plot(history.history['loss']) plt.plot(history.history['val_loss']) plt.title('Model loss') plt.ylabel('Loss') plt.xlabel('Epoch') plt.legend(['Train', 'Test'], loc='upper left') #plt.savefig('figure_dp/loss_trace.eps', format='eps',bbox_inches='tight') plt.show() plt.plot(PredValSet[:,2],true_test_output[:,2],'o', color='blue',markersize=5) #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') plt.show() plt.plot(PredValSet[:,3],true_test_output[:,3],'o',color='blue',markersize=5) #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') plt.show() # predint = model.predict(train_input[:3000]) # trueint = train_output[:3000] # plt.plot(predint[:,3],trueint[:,3],'o', color='blue',markersize=5) # #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') # plt.show()
6,328
25.817797
121
py
LSTM_Covariance
LSTM_Covariance-main/lorenz/simulated_data_generation.py
# -*- coding: utf-8 -*- # generate the trainning set for keras regression import numpy as np from scipy.optimize import fmin from scipy.optimize import fmin_l_bfgs_b #from scipy.optimize import fmin_ncg from scipy.linalg import sqrtm import math from constructB import * from lorentz_attractor import * import matplotlib.pyplot as plt import matplotlib.gridspec as gridspec import time import random import lorentz_attractor import sklearn from sklearn import datasets import os if not os.path.isdir('lorenz_cov_train_v2'): os.makedirs('lorenz_cov_train_v2') ####################################################################### def correlation_from_covariance(covariance): v = np.sqrt(np.diag(covariance)) outer_v = np.outer(v, v) correlation = covariance / outer_v #https://www.mygreatlearning.com/blog/covariance-vs-correlation/ correlation[covariance == 0] = 0 return correlation ###################################################################### #define matrix R by extra-diagonal elements def R_covariance_dim3(r1,r2,r3): M = np.zeros((3,3)) M[0,1] = r1 M[0,2] = r2 M[1,2] = r3 M = M + M.T M += np.eye(3) return M ################################################################################ ####################################################################### def cov_to_cor(M): # from a covariance matrix to its associated correlation matrix inv_diag_M=np.linalg.inv(sqrtm(np.diag(np.diag(M)))) cor_M = np.dot(inv_diag_M, np.dot(M,inv_diag_M)) return cor_M def lorenz_1step(x, y, z, s=10, r=28, b=2.667,dt = 0.001): x_dot, y_dot, z_dot = lorenz(x, y, z) x_next = x + (x_dot * dt) y_next = y + (y_dot * dt) z_next = z + (z_dot * dt) return x_next, y_next, z_next def VAR_3D(xb,Y,H,B,R): #booleen=1 garde la trace xb1=np.copy(xb) xb1.shape=(xb1.size,1) Y.shape = (Y.size,1) dim_x = xb1.size K=np.dot(B,np.dot(np.transpose(H),np.linalg.inv(np.dot(H,np.dot(B,np.transpose(H)))+R))) #matrice de gain, Kalman gain A=np.dot(np.dot((np.eye(dim_x)-np.dot(K,H)),B),np.transpose((np.eye(dim_x)-np.dot(K,H))))+np.dot(np.dot(K,R),np.transpose(K)) #not the kalman filter expression??? vect=np.dot(H,xb1) xa=np.copy(xb1+np.dot(K,(Y-vect))) return xa,A #xa is the new estimated data, A is the new covariance, ################################################################################### ################################################################################### #parameters num_steps = 1000 H = np.array([[1,1,0],[2,0,1],[0,0,3]]) R = 0.001*np.array([[1,0.4,0.1],[0.4,1,0.4],[0.1,0.4,1]]) B =0.01*np.array([[1,0.2,0.],[0.2,1,0.2],[0.,0.2,1]]) #Q = 0.0001*np.eye(3) ################################################################################### #save the trainning set for different R trainning_set = np.zeros((1,num_steps*3+3+4)) ################################################################################### ############################################################################# # true states vector 3 * number_steps xs,ys,zs = lorenz_attractor(s=10, r=28, b=2.667, dt = 0.001, num_steps=1000) x_true = np.zeros((3,num_steps+1)) x_true[0,:] = np.copy(xs) x_true[1,:] = np.copy(ys) x_true[2,:] = np.copy(zs) ############################################################################### for ii in range(2000): if ii%100 ==0: print(ii) # construct observations #========================================================================= #generate x with noise for repetation in range(10): xs,ys,zs = lorenz_attractor(s=10, r=28, b=2.667, dt = 0.001, num_steps = 1000,x0 = 0.+np.random.normal(0, 0.05), y0=1.+np.random.normal(0, 0.05),z0=1.05+np.random.normal(0, 0.05)) x_true = np.zeros((3,num_steps+1)) x_true[0,:] = np.copy(xs) x_true[1,:] = np.copy(ys) x_true[2,:] = np.copy(zs) #========================================================================= y_true = np.zeros((3,num_steps+1)) y_obs = np.zeros((3,num_steps+1)) v = np.random.uniform(0,100.) R = correlation_from_covariance(sklearn.datasets.make_spd_matrix(3)) #SPD covariance r1 = R[0,1] r2 = R[0,2] r3 = R[1,2] R = v*R for i in range(num_steps+1): print("sample time: ",ii) print("iteration time: ",i) x = x_true[:,i] x.shape = (x.size,1) y = np.dot(H,x) #why this is this expression to calculate y? y.shape = (y.size,) y_true[:,i] = y y_noise = np.random.multivariate_normal(np.zeros(3),R) y_noise.shape = (y_noise.size,) y_noise += y y_obs[:,i] = y_noise parameters = np.array([r1,r2,r3,v]) #output for deep learning regression #train_row = np.concatenate((y_obs.ravel(),x_true.ravel())) #input for deep learning #what are the functionalities of these r ->covaraicen! why v is not necessary??? train_row = y_obs.ravel() train_row = np.concatenate((train_row.ravel(),parameters)) train_row.shape = (1,train_row.size) trainning_set = np.concatenate((trainning_set,train_row), axis=0) # if repetation+ii*10==5000: # np.savetxt(f"lorenz_cov_train_v2/trainset_withx_steps1000_test_{10000+repetation+ii*10}.csv", trainning_set, delimiter=",") trainning_set = trainning_set[1:,:] #####################################################################################"" np.savetxt("lorenz_cov_train_v2/trainset_withx_steps1000_test8.csv", trainning_set, delimiter=",")
5,902
34.993902
192
py
LSTM_Covariance
LSTM_Covariance-main/lorenz/constructB.py
# coding: utf8 #construction of matrix B and special H with measure on the boarder import numpy as np import math from scipy.linalg import sqrtm ##def B_Balgovind(n,Sigma,L): ## Gamma = np.identity(n) ## for i in xrange(n): ## for j in xrange(n): ## Gamma[i,j] = ( 1. + abs(i-j)/L)*np.exp(-abs(i-j)/L) ## B = np.dot(Sigma,np.dot(Gamma,Sigma)) ## return B def get_index_2d (dim,n): #get caratesian coordinate j=n % dim j=j/1. #float( i) i=(n-j)/dim return (i,j)# pourquoi float? #identite def identiity(n):#n : taille de vecteur xb B=np.eye(n) return B #Blgovind (Balgovind correlation functions: file:///C:/paper_ml/sibo/Polyphemus-1.2-Guide-2.pdf) def Balgovind(dim,L): sub_B=np.zeros((dim**2,dim**2)) for i in range(dim**2): (a1,b1)=get_index_2d(dim,i) for j in range(dim**2): (a2,b2)=get_index_2d(dim,j) #reprends les donnees caracterisennes r=math.sqrt((a1-a2)**2+(b1-b2)**2) sub_B[i,j]=(1+r/L)*(math.exp(-r/L)) B1=np.concatenate((sub_B, np.zeros((dim**2,dim**2))), axis=1) B2=np.concatenate(( np.zeros((dim**2,dim**2)), sub_B),axis=1) B=np.concatenate((B1,B2), axis=0)# a changer construction matrice B [dim**2*2,dim**2*2] return B def Gaussian(dim,L): sub_B=np.zeros((dim**2,dim**2)) for i in range(dim**2): (a1,b1)=get_index_2d(dim,i) for j in range(dim**2): (a2,b2)=get_index_2d(dim,j) #reprends les donnees caracterisennes r=math.sqrt((a1-a2)**2+(b1-b2)**2) sub_B[i,j]=math.exp(-r**2/(2*L**2)) B1=np.concatenate((sub_B, np.zeros((dim**2,dim**2))), axis=1) B2=np.concatenate(( np.zeros((dim**2,dim**2)), sub_B),axis=1) B=np.concatenate((B1,B2), axis=0)# a changer construction matrice B [dim**2*2,dim**2*2] return B def expontielle(dim,L): sub_B=np.zeros((dim**2,dim**2)) for i in range(dim**2): (a1,b1)=get_index_2d(dim,i) for j in range(dim**2): (a2,b2)=get_index_2d(dim,j) #reprends les donnees caracterisennes r=math.sqrt((a1-a2)**2+(b1-b2)**2) sub_B[i,j]=math.exp(-r/L) B1=np.concatenate((sub_B, np.zeros((dim**2,dim**2))), axis=1) B2=np.concatenate(( np.zeros((dim**2,dim**2)), sub_B),axis=1) B=np.concatenate((B1,B2), axis=0)# a changer construction matrice B [dim**2*2,dim**2*2] return B def bord_M_aleatoire(dimension, proba): M=np.zeros((dimension**2,dimension**2)) for i in range(dimension**2): for j in range(dimension**2): if j % 10==0 or j % 10==9: M[i,j]=np.random.binomial(1, proba) elif j<=9 or j>=90: M[i,j]=np.random.binomial(1, proba) #if there is some problems? because both of them are np.random.binomial???? return M # what are other parameters which are not satisfied with this condition ??? def cov_to_cor(B): inv_diag_B=np.linalg.inv(sqrtm(np.diag(np.diag(B)))) inv_diag_B=np.copy(inv_diag_B.real) cor_B=np.dot(inv_diag_B,np.dot(B,inv_diag_B)) return cor_B ######################################################################## #covariance 1d def Balgovind_1D(dim,L): B=np.zeros((dim,dim)) for i in range(dim): for j in range(dim): r=abs(i-j)*1. B[i,j]=(1+r/L)*(math.exp(-r/L)) return B def expontielle_1D(dim,L): B=np.zeros((dim,dim)) for i in range(dim): for j in range(dim): r=abs(i-j)*1. B[i,j]=math.exp(-r/L) return B def Gaussian_1D(dim,L): B=np.zeros((dim,dim)) for i in range(dim): for j in range(dim): r=abs(i-j)*1. B[i,j]=math.exp(-r**2/(2*L**2)) return B
3,935
32.355932
130
py
LSTM_Covariance
LSTM_Covariance-main/lorenz/lorentz_attractor.py
# -*- coding: utf-8 -*- # lorentz system import numpy as np import time import random import matplotlib.pyplot as plt import itertools import math from constructB import * # This import registers the 3D projection, but is otherwise unused. from mpl_toolkits.mplot3d import Axes3D # noqa: F401 unused import def lorenz(x, y, z, s=10, r=28, b=2.667): ''' Given: x, y, z: a point of interest in three dimensional space s, r, b: parameters defining the lorenz attractor Returns: x_dot, y_dot, z_dot: values of the lorenz attractor's partial derivatives at the point x, y, z; s: sigma r: rho b: beta ''' x_dot = s*(y - x) y_dot = r*x - y - x*z z_dot = x*y - b*z return x_dot, y_dot, z_dot def lorenz_1step(x, y, z, s=10, r=28, b=2.667,dt = 0.001): x_dot, y_dot, z_dot = lorenz(x, y, z) x_next = x + (x_dot * dt) y_next = y + (y_dot * dt) z_next = z + (z_dot * dt) return x_next, y_next, z_next def lorenz_attractor(s=10, r=28, b=2.667, dt = 0.001,num_steps = 1000,x0=0.,y0=1.,z0=1.05 ): # Need one more for the initial values xs = np.empty(num_steps + 1) ys = np.empty(num_steps + 1) zs = np.empty(num_steps + 1) # Set initial values #xs[0], ys[0], zs[0] = (0., 1., 1.05) xs[0] = x0 ys[0] = y0 zs[0] = z0 # Step through "time", calculating the partial derivatives at the current point # and using them to estimate the next point for i in range(num_steps): x_dot, y_dot, z_dot = lorenz(xs[i], ys[i], zs[i]) xs[i + 1] = xs[i] + (x_dot * dt) ys[i + 1] = ys[i] + (y_dot * dt) zs[i + 1] = zs[i] + (z_dot * dt) return xs,ys,zs B = 0.01*np.eye(3) x_noise_initital = np.random.multivariate_normal(np.zeros(3),B) def lorenz_attractor_noisy(s=10, r=28, b=2.667, dt = 0.001,num_steps = 100000, B = 0.01*np.eye(3), Q = 1e-5*np.eye(3)): # Need one more for the initial values xs = np.empty(num_steps + 1) ys = np.empty(num_steps + 1) zs = np.empty(num_steps + 1) # Set initial values xs[0], ys[0], zs[0] = (0., 1., 1.05) #np.random.seed( 10 ) #x_noise = np.random.multivariate_normal(np.zeros(3),B) xs[0] += x_noise_initital[0] ys[0] += x_noise_initital[1] zs[0] += x_noise_initital[2] # Step through "time", calculating the partial derivatives at the current point # and using them to estimate the next point for i in range(num_steps): x_noise_step = np.random.multivariate_normal(np.zeros(3),Q) x_dot, y_dot, z_dot = lorenz(xs[i], ys[i], zs[i]) xs[i + 1] = xs[i] + (x_dot * dt) + x_noise_step[0] ys[i + 1] = ys[i] + (y_dot * dt) + x_noise_step[1] zs[i + 1] = zs[i] + (z_dot * dt) + x_noise_step[2] return xs,ys,zs # Plot if __name__ == '__main__' : # plot for amusing # for i in range(1200,1400): # xs,ys,zs = lorenz_attractor(num_steps = 10*i) # xb,yb,zb = lorenz_attractor(num_steps = 10*i-300) # fig = plt.figure() # ax = fig.gca(projection='3d') # # #ax.plot(xs, ys, zs, lw=0.5) # ax.scatter(xs, ys, zs, lw=0.5) # ax.scatter(xb, yb, zb, 'r',lw=0.35) # ax.set_xlabel("X Axis") # ax.set_ylabel("Y Axis") # ax.set_zlabel("Z Axis") # ax.set_title("Lorenz Attractor") # plt.savefig('tmp_figure/lorenz_catch_'+str(i)+'.png', format='png') # #plt.show() # #plt.pause(3) # plt.close() # # import imageio # images = [] # for i in range(1200,1400): # images.append(imageio.imread('tmp_figure/lorenz_catch_'+str(i)+'.png')) # imageio.mimsave('tmp_figure/lorenz_catch.gif', images) ## # import glob # from PIL import * # # Create the frames # frames = [] # imgs = glob.glob("*.png") # for i in range(2000): # new_frame = Image.open('tmp_figure/lorenz_compose_'+str(i)+'.png') # frames.append(new_frame) # # # Save into a GIF file that loops forever # frames[0].save('tmp_figure/png_to_gif.gif', format='GIF', # append_images=frames[1:], # save_all=True, # duration=300, loop=0) #############################################################################################" # print noisy system xs,ys,zs = lorenz_attractor_noisy(num_steps = 1400) # plot for amusing for i in range(1200,1400): #xs,ys,zs = lorenz_attractor_noisy(num_steps = 10*i) #xb,yb,zb = lorenz_attractor(num_steps = 10*i-300) fig = plt.figure() ax = fig.gca(projection='3d') #ax.plot(xs, ys, zs, lw=0.5) ax.scatter(xs[:i], ys[:i], zs[:i], lw=0.5) #ax.scatter(xb, yb, zb, 'r',lw=0.35) ax.set_xlabel("X Axis") ax.set_ylabel("Y Axis") ax.set_zlabel("Z Axis") ax.set_title("Lorenz Attractor") plt.savefig('tmp_figure/lorenz_noisy_'+str(i)+'.png', format='png') #plt.show() #plt.pause(3) plt.close() import imageio images = [] for i in range(1200,1400): images.append(imageio.imread('tmp_figure/lorenz_noisy_'+str(i)+'.png')) imageio.mimsave('tmp_figure/lorenz_noisy.gif', images)
5,328
30.720238
119
py
LSTM_Covariance
LSTM_Covariance-main/lorenz/lorenz_lstm200.py
# -*- coding: utf-8 -*- """ Created on Wed Jan 6 13:39:58 2021 @author: siboc """ import numpy as np import scipy import math import matplotlib.pyplot as plt from keras.models import Sequential from keras.layers import Dense from sklearn.metrics import r2_score import tensorflow as tf import keras.backend as K # check scikit-learn version # check scikit-learn version import sklearn from sklearn.linear_model import LinearRegression from sklearn.datasets import make_regression from sklearn.neighbors import KNeighborsRegressor from sklearn.tree import DecisionTreeRegressor import pandas as pd def data_set_order(file): train_data = np.array(pd.read_csv(file))[:-2,:] r0=train_data[:,:1001][:,:201] r1=train_data[:,1001:2002][:,:201] r2=train_data[:,2002:3003][:,:201] r3=train_data[:,3003:] r3[:,-1]=r3[:,-1]/100 train_data=np.insert(r0,[i+1 for i in range(r0.shape[1])],r1,axis=1) train_data=np.insert(train_data,[(i+1)*2 for i in range(int(train_data.shape[1]/2))],r2,axis=1) train_data=np.concatenate((train_data,r3),axis=1) return train_data #input train_data = data_set_order('lorenz_cov_train_v2/trainset_withx_steps1000_1.csv') print("train_data shape: ",train_data.shape) # train_data1 = data_set_order('lorenz_cov_train/trainset_withx_repeat10bis3.csv') # print("train_data1 shape: ",train_data1.shape) # train_data2 = data_set_order('lorenz_cov_train/trainset_withx_repeat10bis4.csv') # print("train_data2 shape: ",train_data2.shape) # train_data3 = data_set_order('lorenz_cov_train/trainset_withx_repeat10bis5.csv') # print("train_data3 shape: ",train_data3.shape) # train_data4 = data_set_order('lorenz_cov_train/trainset_withx_repeat10bis6.csv') # print("train_data4 shape: ",train_data4.shape) # train_data5 = data_set_order('lorenz_cov_train/trainset_withx_repeat10bis7.csv') # print("train_data5 shape: ",train_data5.shape) # train_data6 = data_set_order('lorenz_cov_train/trainset_withx_repeat10bis8.csv') # print("train_data6 shape: ",train_data6.shape) #size: num_steps*3,r1,r2,r3,v ######################################################################################### #train_data = np.array(pd.read_csv('data_1000steps/trainset_withx_1000steps.csv')) # # #train_data1 = np.array(pd.read_csv('data_1000steps/trainset_withx_1000stepsbis1.csv')) # #train_data2 = np.array(pd.read_csv('data_1000steps/trainset_withx_1000stepsbis2.csv')) # #train_data3 = np.array(pd.read_csv('data_1000steps/trainset_withx_1000stepsbis3.csv')) # train_data = np.concatenate((train_data6,train_data5),axis = 0) # train_data = np.concatenate((train_data,train_data4),axis = 0) # train_data = np.concatenate((train_data,train_data3),axis = 0) # train_data = np.concatenate((train_data,train_data2),axis = 0) # train_data = np.concatenate((train_data,train_data1),axis = 0) # train_data = np.concatenate((train_data,train_data0),axis = 0) # train_data=train_data[:120000,:] #weightstrain_data[:,604:] np.random.shuffle(train_data ) input_data = train_data[:,0:603] output_data = train_data[:,603:] ######################################################################## train_part = 0.97 threshold = int(train_part*train_data.shape[0]) ########################################################################## train_input = input_data[:threshold] print("input_data shape: ",input_data.shape) train_output = output_data[:threshold] print("output_data shape: ",output_data.shape) test_input = input_data [threshold:] true_test_output = output_data[threshold:] X1 = train_input Y1 = train_output X2 = test_input #Y2 = ValidationSet_Y ############################################################################ #def my_loss_fn(y_true, y_pred): # # return K.mean(K.abs(y_true - y_pred) * weight) # ======================================================================================== from keras.layers import LSTM,Dropout from keras.layers import TimeDistributed from keras.callbacks import EarlyStopping from keras.callbacks import ModelCheckpoint from keras.optimizers import Adam # save data import os import json import pickle if not os.path.isdir('save_data_v2'): os.makedirs('save_data_v2') hidden_size=200 input_sample=input_data.shape[0] #for one sample output_sample=output_data.shape[0] input_data=input_data.reshape(input_sample,201,3) #201 is the time steps in data_generation output_data=output_data.reshape(output_sample,4) use_dropout=True model = Sequential() model.add(LSTM(hidden_size,input_shape=(201,3))) model.add(Dense(4)) # opt = Adam(lr=0.0001) model.compile(loss='mean_squared_error', optimizer='Adam', metrics=['mae']) print(model.summary()) es=EarlyStopping(monitor='val_loss',mode='min',verbose=1,patience=50) # modelcheckpoint mc=ModelCheckpoint('save_data_v2/sequentiallstm200_ing.h5',monitor='val_loss',mode='min',save_best_only=True,verbose=1) history=model.fit(input_data, output_data, validation_split=0.1, epochs=100, batch_size=128, verbose=1,callbacks=[es,mc]) # model.save('save_data/sequentiallstm2') model.save('save_data_v2/sequentiallstm200_ing_f.h5') # https://stackoverflow.com/a/44674337/10349608 with open('save_data_v2/sequentiallstm200_ing_history.pickle', 'wb') as file_his: pickle.dump(history.history, file_his) # Calculate predictions PredTestSet = model.predict(X1.reshape(X1.shape[0],201,3)) PredValSet = model.predict(X2.reshape(X2.shape[0],201,3)) plt.plot(history.history['loss']) plt.plot(history.history['val_loss']) plt.title('Model loss') plt.ylabel('Loss') plt.xlabel('Epoch') plt.legend(['Train', 'Test'], loc='upper left') #plt.savefig('figure_dp/loss_trace.eps', format='eps',bbox_inches='tight') plt.show() plt.plot(PredValSet[:,2],true_test_output[:,2],'o', color='blue',markersize=5) #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') plt.show() plt.plot(PredValSet[:,3],true_test_output[:,3],'o',color='blue',markersize=5) #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') plt.show() # predint = model.predict(train_input[:3000]) # trueint = train_output[:3000] # plt.plot(predint[:,3],trueint[:,3],'o', color='blue',markersize=5) # #plt.plot(list(range(0,1,0.1)),list(range(0,1,0.1)),'k') # plt.show()
6,201
25.618026
121
py
PMEmo
PMEmo-master/features.py
#! usr/bin/env python3 # -*- coding: utf-8 -*- ''' This features.py is used to extract audio features based on openSIMLE. Require: openSMILE-2.2rc1 OpenSMILE only support audios in WAV format, so before using this script you could transform MP3s into WAVs by transformat.sh. ''' __author__ = 'huizhang' import csv import os import shutil import subprocess from math import floor import numpy as np def extract_all_wav_feature(wavdir, distfile, opensmiledir): '''Extract 6373-dimension static features into one dist file. Args: wavdir: Path to audios in WAV format. distfile: Path of distfile. opensmiledir: Path to opensimle project root. Returns: Distfile containing 6373-dimension static features of all the WAVs. ''' SMILExtract = os.path.join(opensmiledir,"SMILExtract") config_file = os.path.join(opensmiledir,"config", "IS13_ComParE.conf") if os.path.exists(distfile): os.remove(distfile) wav = [f for f in os.listdir(wavdir) if f[-4:] == ".wav"] for w in wav: wavpath = os.path.join(wavdir,w) subprocess.check_call([SMILExtract, "-C", config_file, "-I", wavpath, "-O", distfile, "-instname", w]) def extract_frame_feature(wavdir, distdir, opensmiledir): '''Extract lld features in frame size: 60ms, step size: 10ms. Args: wavdir: Path to audios in WAV format. distdir: Path of distdir. opensmiledir: Path to opensimle project root. Returns: Distfiles containing lld features for each WAV. ''' SMILExtract = os.path.join(opensmiledir,"SMILExtract") config_file = os.path.join(opensmiledir,"config", "IS13_ComParE_lld.conf") if os.path.exists(distdir): shutil.rmtree(distdir) os.mkdir(distdir) wav = [f for f in os.listdir(wavdir) if f[-4:] == ".wav"] for w in wav: wavpath = os.path.join(wavdir,w) distfile = os.path.join(distdir,w[:-4]+".csv") subprocess.check_call([SMILExtract, "-C", config_file, "-I", wavpath, "-O", distfile]) def process_dynamic_feature(llddir, distdir, all_songs_distfile, delimiter=";"): '''Obtain dynamic features in window size: 1s, shift size: 0.5s. Args: llddir: Path to lld feature files. distdir: Path of distdir. all_songs_distfile: Path of distfile. delimiter: csv delimiter in lld feature files, default=';'. Returns: Distfiles containing 260-dimension dynamic features all WAVs. ''' if os.path.exists(distdir): shutil.rmtree(distdir) os.mkdir(distdir) # names of features headers = ['musicId', 'frameTime', 'F0final_sma_mean', 'voicingFinalUnclipped_sma_mean', 'jitterLocal_sma_mean', 'jitterDDP_sma_mean', 'shimmerLocal_sma_mean', 'logHNR_sma_mean', 'audspec_lengthL1norm_sma_mean', 'audspecRasta_lengthL1norm_sma_mean', 'pcm_RMSenergy_sma_mean', 'pcm_zcr_sma_mean', 'audSpec_Rfilt_sma[0]_mean', 'audSpec_Rfilt_sma[1]_mean', 'audSpec_Rfilt_sma[2]_mean', 'audSpec_Rfilt_sma[3]_mean', 'audSpec_Rfilt_sma[4]_mean', 'audSpec_Rfilt_sma[5]_mean', 'audSpec_Rfilt_sma[6]_mean', 'audSpec_Rfilt_sma[7]_mean', 'audSpec_Rfilt_sma[8]_mean', 'audSpec_Rfilt_sma[9]_mean', 'audSpec_Rfilt_sma[10]_mean', 'audSpec_Rfilt_sma[11]_mean', 'audSpec_Rfilt_sma[12]_mean', 'audSpec_Rfilt_sma[13]_mean', 'audSpec_Rfilt_sma[14]_mean', 'audSpec_Rfilt_sma[15]_mean', 'audSpec_Rfilt_sma[16]_mean', 'audSpec_Rfilt_sma[17]_mean', 'audSpec_Rfilt_sma[18]_mean', 'audSpec_Rfilt_sma[19]_mean', 'audSpec_Rfilt_sma[20]_mean', 'audSpec_Rfilt_sma[21]_mean', 'audSpec_Rfilt_sma[22]_mean', 'audSpec_Rfilt_sma[23]_mean', 'audSpec_Rfilt_sma[24]_mean', 'audSpec_Rfilt_sma[25]_mean', 'pcm_fftMag_fband250-650_sma_mean', 'pcm_fftMag_fband1000-4000_sma_mean', 'pcm_fftMag_spectralRollOff25.0_sma_mean', 'pcm_fftMag_spectralRollOff50.0_sma_mean', 'pcm_fftMag_spectralRollOff75.0_sma_mean', 'pcm_fftMag_spectralRollOff90.0_sma_mean', 'pcm_fftMag_spectralFlux_sma_mean', 'pcm_fftMag_spectralCentroid_sma_mean', 'pcm_fftMag_spectralEntropy_sma_mean', 'pcm_fftMag_spectralVariance_sma_mean', 'pcm_fftMag_spectralSkewness_sma_mean', 'pcm_fftMag_spectralKurtosis_sma_mean', 'pcm_fftMag_spectralSlope_sma_mean', 'pcm_fftMag_psySharpness_sma_mean', 'pcm_fftMag_spectralHarmonicity_sma_mean', 'pcm_fftMag_mfcc_sma[1]_mean', 'pcm_fftMag_mfcc_sma[2]_mean', 'pcm_fftMag_mfcc_sma[3]_mean', 'pcm_fftMag_mfcc_sma[4]_mean', 'pcm_fftMag_mfcc_sma[5]_mean', 'pcm_fftMag_mfcc_sma[6]_mean', 'pcm_fftMag_mfcc_sma[7]_mean', 'pcm_fftMag_mfcc_sma[8]_mean', 'pcm_fftMag_mfcc_sma[9]_mean', 'pcm_fftMag_mfcc_sma[10]_mean', 'pcm_fftMag_mfcc_sma[11]_mean', 'pcm_fftMag_mfcc_sma[12]_mean', 'pcm_fftMag_mfcc_sma[13]_mean', 'pcm_fftMag_mfcc_sma[14]_mean', 'F0final_sma_de_mean', 'voicingFinalUnclipped_sma_de_mean', 'jitterLocal_sma_de_mean', 'jitterDDP_sma_de_mean', 'shimmerLocal_sma_de_mean', 'logHNR_sma_de_mean', 'audspec_lengthL1norm_sma_de_mean', 'audspecRasta_lengthL1norm_sma_de_mean', 'pcm_RMSenergy_sma_de_mean', 'pcm_zcr_sma_de_mean', 'audSpec_Rfilt_sma_de[0]_mean', 'audSpec_Rfilt_sma_de[1]_mean', 'audSpec_Rfilt_sma_de[2]_mean', 'audSpec_Rfilt_sma_de[3]_mean', 'audSpec_Rfilt_sma_de[4]_mean', 'audSpec_Rfilt_sma_de[5]_mean', 'audSpec_Rfilt_sma_de[6]_mean', 'audSpec_Rfilt_sma_de[7]_mean', 'audSpec_Rfilt_sma_de[8]_mean', 'audSpec_Rfilt_sma_de[9]_mean', 'audSpec_Rfilt_sma_de[10]_mean', 'audSpec_Rfilt_sma_de[11]_mean', 'audSpec_Rfilt_sma_de[12]_mean', 'audSpec_Rfilt_sma_de[13]_mean', 'audSpec_Rfilt_sma_de[14]_mean', 'audSpec_Rfilt_sma_de[15]_mean', 'audSpec_Rfilt_sma_de[16]_mean', 'audSpec_Rfilt_sma_de[17]_mean', 'audSpec_Rfilt_sma_de[18]_mean', 'audSpec_Rfilt_sma_de[19]_mean', 'audSpec_Rfilt_sma_de[20]_mean', 'audSpec_Rfilt_sma_de[21]_mean', 'audSpec_Rfilt_sma_de[22]_mean', 'audSpec_Rfilt_sma_de[23]_mean', 'audSpec_Rfilt_sma_de[24]_mean', 'audSpec_Rfilt_sma_de[25]_mean', 'pcm_fftMag_fband250-650_sma_de_mean', 'pcm_fftMag_fband1000-4000_sma_de_mean', 'pcm_fftMag_spectralRollOff25.0_sma_de_mean', 'pcm_fftMag_spectralRollOff50.0_sma_de_mean', 'pcm_fftMag_spectralRollOff75.0_sma_de_mean', 'pcm_fftMag_spectralRollOff90.0_sma_de_mean', 'pcm_fftMag_spectralFlux_sma_de_mean', 'pcm_fftMag_spectralCentroid_sma_de_mean', 'pcm_fftMag_spectralEntropy_sma_de_mean', 'pcm_fftMag_spectralVariance_sma_de_mean', 'pcm_fftMag_spectralSkewness_sma_de_mean', 'pcm_fftMag_spectralKurtosis_sma_de_mean', 'pcm_fftMag_spectralSlope_sma_de_mean', 'pcm_fftMag_psySharpness_sma_de_mean', 'pcm_fftMag_spectralHarmonicity_sma_de_mean', 'pcm_fftMag_mfcc_sma_de[1]_mean', 'pcm_fftMag_mfcc_sma_de[2]_mean', 'pcm_fftMag_mfcc_sma_de[3]_mean', 'pcm_fftMag_mfcc_sma_de[4]_mean', 'pcm_fftMag_mfcc_sma_de[5]_mean', 'pcm_fftMag_mfcc_sma_de[6]_mean', 'pcm_fftMag_mfcc_sma_de[7]_mean', 'pcm_fftMag_mfcc_sma_de[8]_mean', 'pcm_fftMag_mfcc_sma_de[9]_mean', 'pcm_fftMag_mfcc_sma_de[10]_mean', 'pcm_fftMag_mfcc_sma_de[11]_mean', 'pcm_fftMag_mfcc_sma_de[12]_mean', 'pcm_fftMag_mfcc_sma_de[13]_mean', 'pcm_fftMag_mfcc_sma_de[14]_mean', 'F0final_sma_std', 'voicingFinalUnclipped_sma_std', 'jitterLocal_sma_std', 'jitterDDP_sma_std', 'shimmerLocal_sma_std', 'logHNR_sma_std', 'audspec_lengthL1norm_sma_std', 'audspecRasta_lengthL1norm_sma_std', 'pcm_RMSenergy_sma_std', 'pcm_zcr_sma_std', 'audSpec_Rfilt_sma[0]_std', 'audSpec_Rfilt_sma[1]_std', 'audSpec_Rfilt_sma[2]_std', 'audSpec_Rfilt_sma[3]_std', 'audSpec_Rfilt_sma[4]_std', 'audSpec_Rfilt_sma[5]_std', 'audSpec_Rfilt_sma[6]_std', 'audSpec_Rfilt_sma[7]_std', 'audSpec_Rfilt_sma[8]_std', 'audSpec_Rfilt_sma[9]_std', 'audSpec_Rfilt_sma[10]_std', 'audSpec_Rfilt_sma[11]_std', 'audSpec_Rfilt_sma[12]_std', 'audSpec_Rfilt_sma[13]_std', 'audSpec_Rfilt_sma[14]_std', 'audSpec_Rfilt_sma[15]_std', 'audSpec_Rfilt_sma[16]_std', 'audSpec_Rfilt_sma[17]_std', 'audSpec_Rfilt_sma[18]_std', 'audSpec_Rfilt_sma[19]_std', 'audSpec_Rfilt_sma[20]_std', 'audSpec_Rfilt_sma[21]_std', 'audSpec_Rfilt_sma[22]_std', 'audSpec_Rfilt_sma[23]_std', 'audSpec_Rfilt_sma[24]_std', 'audSpec_Rfilt_sma[25]_std', 'pcm_fftMag_fband250-650_sma_std', 'pcm_fftMag_fband1000-4000_sma_std', 'pcm_fftMag_spectralRollOff25.0_sma_std', 'pcm_fftMag_spectralRollOff50.0_sma_std', 'pcm_fftMag_spectralRollOff75.0_sma_std', 'pcm_fftMag_spectralRollOff90.0_sma_std', 'pcm_fftMag_spectralFlux_sma_std', 'pcm_fftMag_spectralCentroid_sma_std', 'pcm_fftMag_spectralEntropy_sma_std', 'pcm_fftMag_spectralVariance_sma_std', 'pcm_fftMag_spectralSkewness_sma_std', 'pcm_fftMag_spectralKurtosis_sma_std', 'pcm_fftMag_spectralSlope_sma_std', 'pcm_fftMag_psySharpness_sma_std', 'pcm_fftMag_spectralHarmonicity_sma_std', 'pcm_fftMag_mfcc_sma[1]_std', 'pcm_fftMag_mfcc_sma[2]_std', 'pcm_fftMag_mfcc_sma[3]_std', 'pcm_fftMag_mfcc_sma[4]_std', 'pcm_fftMag_mfcc_sma[5]_std', 'pcm_fftMag_mfcc_sma[6]_std', 'pcm_fftMag_mfcc_sma[7]_std', 'pcm_fftMag_mfcc_sma[8]_std', 'pcm_fftMag_mfcc_sma[9]_std', 'pcm_fftMag_mfcc_sma[10]_std', 'pcm_fftMag_mfcc_sma[11]_std', 'pcm_fftMag_mfcc_sma[12]_std', 'pcm_fftMag_mfcc_sma[13]_std', 'pcm_fftMag_mfcc_sma[14]_std', 'F0final_sma_de_std', 'voicingFinalUnclipped_sma_de_std', 'jitterLocal_sma_de_std', 'jitterDDP_sma_de_std', 'shimmerLocal_sma_de_std', 'logHNR_sma_de_std', 'audspec_lengthL1norm_sma_de_std', 'audspecRasta_lengthL1norm_sma_de_std', 'pcm_RMSenergy_sma_de_std', 'pcm_zcr_sma_de_std', 'audSpec_Rfilt_sma_de[0]_std', 'audSpec_Rfilt_sma_de[1]_std', 'audSpec_Rfilt_sma_de[2]_std', 'audSpec_Rfilt_sma_de[3]_std', 'audSpec_Rfilt_sma_de[4]_std', 'audSpec_Rfilt_sma_de[5]_std', 'audSpec_Rfilt_sma_de[6]_std', 'audSpec_Rfilt_sma_de[7]_std', 'audSpec_Rfilt_sma_de[8]_std', 'audSpec_Rfilt_sma_de[9]_std', 'audSpec_Rfilt_sma_de[10]_std', 'audSpec_Rfilt_sma_de[11]_std', 'audSpec_Rfilt_sma_de[12]_std', 'audSpec_Rfilt_sma_de[13]_std', 'audSpec_Rfilt_sma_de[14]_std', 'audSpec_Rfilt_sma_de[15]_std', 'audSpec_Rfilt_sma_de[16]_std', 'audSpec_Rfilt_sma_de[17]_std', 'audSpec_Rfilt_sma_de[18]_std', 'audSpec_Rfilt_sma_de[19]_std', 'audSpec_Rfilt_sma_de[20]_std', 'audSpec_Rfilt_sma_de[21]_std', 'audSpec_Rfilt_sma_de[22]_std', 'audSpec_Rfilt_sma_de[23]_std', 'audSpec_Rfilt_sma_de[24]_std', 'audSpec_Rfilt_sma_de[25]_std', 'pcm_fftMag_fband250-650_sma_de_std', 'pcm_fftMag_fband1000-4000_sma_de_std', 'pcm_fftMag_spectralRollOff25.0_sma_de_std', 'pcm_fftMag_spectralRollOff50.0_sma_de_std', 'pcm_fftMag_spectralRollOff75.0_sma_de_std', 'pcm_fftMag_spectralRollOff90.0_sma_de_std', 'pcm_fftMag_spectralFlux_sma_de_std', 'pcm_fftMag_spectralCentroid_sma_de_std', 'pcm_fftMag_spectralEntropy_sma_de_std', 'pcm_fftMag_spectralVariance_sma_de_std', 'pcm_fftMag_spectralSkewness_sma_de_std', 'pcm_fftMag_spectralKurtosis_sma_de_std', 'pcm_fftMag_spectralSlope_sma_de_std', 'pcm_fftMag_psySharpness_sma_de_std', 'pcm_fftMag_spectralHarmonicity_sma_de_std', 'pcm_fftMag_mfcc_sma_de[1]_std', 'pcm_fftMag_mfcc_sma_de[2]_std', 'pcm_fftMag_mfcc_sma_de[3]_std', 'pcm_fftMag_mfcc_sma_de[4]_std', 'pcm_fftMag_mfcc_sma_de[5]_std', 'pcm_fftMag_mfcc_sma_de[6]_std', 'pcm_fftMag_mfcc_sma_de[7]_std', 'pcm_fftMag_mfcc_sma_de[8]_std', 'pcm_fftMag_mfcc_sma_de[9]_std', 'pcm_fftMag_mfcc_sma_de[10]_std', 'pcm_fftMag_mfcc_sma_de[11]_std', 'pcm_fftMag_mfcc_sma_de[12]_std', 'pcm_fftMag_mfcc_sma_de[13]_std', 'pcm_fftMag_mfcc_sma_de[14]_std'] window = 1 overlap = 0.5 llds = [f for f in os.listdir(llddir) if f[-4:] == ".csv"] all_dynamic_features = [] all_musicId = [] for lld in llds: musicId = [] lldpath = os.path.join(llddir,lld) single_song_distfile = os.path.join(distdir,lld) dynamic_features = _compute_feature_with_window_and_overlap(lldpath, window, overlap, delimiter) for i in range(len(dynamic_features)): musicId.append(lld[:-4]) _write_features_to_csv(headers, musicId, dynamic_features, single_song_distfile) all_musicId += musicId all_dynamic_features += dynamic_features _write_features_to_csv(headers, all_musicId, all_dynamic_features, all_songs_distfile) def _compute_feature_with_window_and_overlap(lldpath, window, overlap, delimiter): '''Compute the mean and std for frame-wise features in window size: 1s, shift size: 0.5s.''' fs = 0.01 num_in_new_frame = floor(overlap/fs) num_in_window = floor(window/fs) # load the features from disk all_frame = [] with open(lldpath) as f: reader = csv.reader(f,delimiter=delimiter) next(reader) for row in reader: frame_feature = [] for i in range(len(row)-1): #旧的frametime不用记录 frame_feature.append(float(row[i+1])) all_frame.append(frame_feature) # compute new number of frames new_num_of_frame = floor(len(all_frame)/num_in_new_frame) all_new_frame = [] # compute mean and std in each window as the feature corresponding to the frame. for i in range(new_num_of_frame): start_index = num_in_new_frame * i new_frame_array = np.array(all_frame[start_index:start_index+num_in_window]) mean_llds = np.mean(new_frame_array,axis=0) std_llds = np.std(new_frame_array,axis=0) new_frametime = i * overlap new_frame = [new_frametime] + mean_llds.tolist() + std_llds.tolist() all_new_frame.append(new_frame) return all_new_frame def _write_features_to_csv(headers, musicIds, contents, distfile): '''Write all the features into one file, and add the last column as the annotation value''' with open(distfile,"w") as newfile: writer = csv.writer(newfile) writer.writerow(headers + ["class"]) for i in range(len(contents)): writer.writerow([musicIds[i]] + contents[i] + ["?"]) if __name__ == "__main__": wavdir ="/Path/to/WAVs" opensmiledir = "/Path/to/openSMILE-2.1.0" static_distfile = "static_features.arff" lld_distdir = "IS13features_lld" dynamic_distdir = "dynamic_features" all_dynamic_distfile = "dynamic_features.csv" delimiter = ";" extract_all_wav_feature(wavdir,static_distfile,opensmiledir) extract_frame_feature(wavdir,lld_distdir,opensmiledir) process_dynamic_feature(lld_distdir,dynamic_distdir,all_dynamic_distfile,delimiter)
14,123
81.116279
8,566
py
ExplainableAIImageMeasures
ExplainableAIImageMeasures-main/setup.py
import pathlib from setuptools import setup # The directory containing this file HERE = pathlib.Path(__file__).parent # The text of the README file README = (HERE / "README.md").read_text() # This call to setup() does all the work setup( name="explainable_ai_image_measures", version="1.0.1", description="Compute IAUC, DAUC, IROF scores to measure quality of image attributions", long_description=README, long_description_content_type="text/markdown", url="https://github.com/Meier-Johannes/ExplainableAIImageMeasures", author="Johannes Meier", author_email="johannes-michael.meier@student.uni-tuebingen.de", license="MIT", classifiers=[ "License :: OSI Approved :: MIT License", "Programming Language :: Python :: 3", "Programming Language :: Python :: 3.8", ], packages=["explainable_ai_image_measures"], include_package_data=True, install_requires=["numpy", "torch", "scikit-image", "scikit-learn"], )
989
32
91
py
ExplainableAIImageMeasures
ExplainableAIImageMeasures-main/explainable_ai_image_measures/scoring_metric.py
import numpy as np import torch from sklearn.metrics import auc import torch.nn.functional as F from explainable_ai_image_measures.irof import IrofDataset from explainable_ai_image_measures.pixel_relevancy import PixelRelevancyDataset class Measures: def __init__(self, model, batch_size=64, irof_segments=40, irof_sigma=5, pixel_package_size=1, normalize=True, clip01=False, baseline_color=None): """ Parametrize the future measurements model: PyTorch model batch_size: During each iteration batch_size number of images will be sent through the network simultaneously irof_segments: Maximum number of slic segments, that we want to use for measuring. Only relevant if you compute IROF later irof_sigma: Parameter used in the slic algorithm pixel_package_size: E.g. for imagenet you may have 224*224=50,176 pixels. Therefore sending 50,176 pixels through the network may lead to too much overhead. Instead you can also remove / add blocks of pixels to speed up the computation. Only relevant for IAUC, DAUC normalize: With activated normalization the new probabilities are divided by the probabiilties of the old image. This allows the comparison of attributions independent of how sure the network is for the original image. Activating normalization is highly encouraged if comparing attributions across several images. clip01: Clips the computed probabilities between [0, 1]. This is only relevant for normalize=True. In some cases the probabilities after e.g. removing parts of the original image may be higher than before. E.g. for IROF this could theoretically lead to negative scores. If you want to prohibit this, activate clip01. Note that the clipping clips each individual score. Indirectly you also ensure that the final score is within [0,1] baseline_color: For IROF and DAUC we iteratively remove parts of the image and replace it by the baseline color as specified here. For IAUC we start with an image consisting only of the baseline_color. By default the mean color is used. """ self.model = model self.batch_size = batch_size self.irof_segments = irof_segments self.irof_sigma = irof_sigma self.pixel_package_size = pixel_package_size self.normalize = normalize self.clip01 = clip01 self.baseline_color = baseline_color def _calc_probs(self, image_batch, label): probs = F.softmax(self.model(image_batch), dim=1) return probs[:, label] def _calc_single_score(self, scoring_dataset, label): probs = [] with torch.no_grad(): for j, img_batch in enumerate(scoring_dataset): probs += [self._calc_probs(img_batch, label)] probs = torch.cat(probs).flatten() if self.normalize: probs = probs[:-1] / probs[-1] else: probs = probs[:-1] if self.clip01: probs = torch.clamp(probs, 0, 1) probs = scoring_dataset.postprocess_scores(probs) x = np.arange(0, len(probs)) y = probs.detach().cpu().numpy() score = auc(x, y) / len(probs) return score, probs.detach() def _assert_check(self, image, attribution): assert(len(image.shape) == 3) assert(image.shape[1:] == attribution.shape) if self.baseline_color is not None: assert(len(self.baseline_color.shape) == 1) assert(len(self.baseline_color) == image.shape[0]) def compute_IAUC(self, image, attribution, label): """ Computes IAUC for a single image and attribution image: Torch.FloatTensor(color_channel, width, height) attribution: Torch.FloatTensor(width, height) label: Label of the attribution """ self._assert_check(image, attribution) with torch.no_grad(): dataset = PixelRelevancyDataset( image, attribution, True, self.batch_size, self.pixel_package_size, image.device, self.baseline_color ) return self._calc_single_score(dataset, label) def compute_DAUC(self, image, attribution, label): """ Computes DAUC for a single image and attribution image: Torch.FloatTensor(color_channel, width, height) attribution: Torch.FloatTensor(width, height) label: Label of the attribution """ self._assert_check(image, attribution) with torch.no_grad(): dataset = PixelRelevancyDataset( image, attribution, False, self.batch_size, self.pixel_package_size, image.device, self.baseline_color ) return self._calc_single_score(dataset, label) def compute_IROF(self, image, attribution, label): """ Computes IROF for a single image and attribution image: Torch.FloatTensor(color_channel, width, height) attribution: Torch.FloatTensor(width, height) label: Label of the attribution """ self._assert_check(image, attribution) with torch.no_grad(): dataset = IrofDataset( image, attribution, self.batch_size, self.irof_segments, self.irof_sigma, image.device, self.baseline_color ) return self._calc_single_score(dataset, label) def compute_batch(self, images, attributions, labels, IROF=True, IAUC=True, DAUC=True): """ Computes the batch for many images and allows multiple attributions per image. image: Torch.FloatTensor(nr_images, color_channel, width, height) attribution: (nr_images, nr_attributions_per_image, width, height) labels: Tuple / Array / Tensor of Int IROF: Defines, whether IROF is computed IAUC: Defines, whether IAUC is computed DAUC: Defines, whether DAUC is computed """ assert(len(images) == len(attributions)) assert(len(images) == len(labels)) functions = dict() if IROF: functions["IROF"] = self.compute_IROF if IAUC: functions["IAUC"] = self.compute_IAUC if DAUC: functions["DAUC"] = self.compute_DAUC if len(functions) == 0: return None result = dict() for method in functions: scores = torch.zeros(attributions.shape[0:2]) probs = [] for img_id in range(len(images)): probs.append([]) for attr_id in range(len(attributions[img_id])): score, prob = functions[method]( images[img_id], attributions[img_id, attr_id], labels[img_id] ) scores[img_id, attr_id] = score probs[-1].append(prob) probs[-1] = torch.stack(probs[-1]) result[method] = (scores, probs) return result
7,637
37.771574
120
py
ExplainableAIImageMeasures
ExplainableAIImageMeasures-main/explainable_ai_image_measures/pixel_manipulation.py
import torch from torch.utils.data import Dataset import abc class PixelManipulationBase(Dataset): """ Requires that self._pixel_batches is defined in the constructor """ def __init__(self, image, attribution, insert, batch_size, device, baseline_color): self._image = image self._batch_size = batch_size self._insert = insert self._attribution = attribution self._device = device self._baseline_color = baseline_color self._pixel_batches = None # Expected to be set by child class self._temp_image = None # Expected to be set by child class if self._baseline_color is None: self._baseline_color = torch.mean(image.reshape(self.color_channels, self.width, self.height), dim=(1, 2)) self._temp_baseline = None self._temp_image = None @abc.abstractmethod def generate_pixel_batches(self): return def generate_initial_image(self): if self._insert: # Create a new image of original width & height with every pixel set to the baseline color self._temp_image = ( self._baseline_color.view(self.color_channels, 1) .repeat(1, self.width * self.height) ) else: self._temp_image = self._image self._temp_image = self._temp_image.flatten() def generate_temp_baseline(self): # Keep baseline image multiply times to avoid repeat the generation every iteration self._temp_baseline = ( self._baseline_color.view(self.color_channels, 1).repeat(1, self.nr_pixels).flatten() ) def __len__(self): return len(self._pixel_batches) @property def color_channels(self): return self._image.shape[0] @property def width(self): return self._image.shape[1] @property def height(self): return self._image.shape[2] def _get_batch_size(self, index): return len(self._pixel_batches[index]) def _index_shift(self, matrix, add_per_row): # Adds (i-1) * add_per_row to every cell in row i. Nothing for row 1 factor = add_per_row * torch.diag(torch.arange(0, len(matrix), device=self._device)).float() new_matrix = factor @ torch.ones_like(matrix, device=self._device).float() + matrix return new_matrix.long() @abc.abstractmethod def _gen_indices(self, index): return @property def nr_pixels(self): return self.width * self.height def _color_channel_shift(self, indices): # For every color channel shift by nr_pixels return torch.stack( [indices + i*self.nr_pixels for i in range(self.color_channels)] ).to(self._device).T.flatten() def _batch_shift(self, indices, pixel_per_image): # Depending on pixel_per_image create an array of the following form: # [0 0 1 1 1 2 2 2 2 2 2] # How often a number is repeated depends on pixel_per_image nr_pixels_cum = torch.cumsum(pixel_per_image, 0) image_indices = [torch.Tensor(nr_pixels_cum[i]*[i]) for i in range(len(nr_pixels_cum))] image_indices = torch.cat(image_indices).to(self._device).long() # Multiply it with the total number of data points per image image_indices_shift = image_indices * self.color_channels * self.nr_pixels # Expand the shift for each color channel nr_man_pixels = int(len(indices) / self.color_channels) image_indices_shift = image_indices_shift.reshape(-1, 1).expand(nr_man_pixels, self.color_channels) image_indices_shift = image_indices_shift.flatten() # Shift the original indices batch_indices = indices + image_indices_shift return batch_indices @abc.abstractmethod def _get_fake_image_size(self): return @abc.abstractmethod def postprocess_scores(self, y): return y def __getitem__(self, index): """ Returns batch_size of images, where the most important pixels have been removed / added Important: Call the method with consecutive index values! """ batch_size = self._get_batch_size(index) # Start with the image of the last run # Create batch of image [batch_size, color_channels x width x height] image_batch = self._temp_image.view(1, -1).repeat(batch_size, 1).flatten() # Get the indices that need to be modified from the image of the last run # template_indices = not unique, batch_indices = unique template_indices, batch_indices = self._gen_indices(index) if index == self.__len__() - 1: # Only in the last run: Ensure that there is no problem for the original image # Therefore remove the fake indices as added in the constructor template_indices = template_indices[ : -self._get_fake_image_size() * self.color_channels * batch_size ] batch_indices = batch_indices[ : -self._get_fake_image_size() * self.color_channels * batch_size ] # Modify the pixels if self._insert: image_batch[batch_indices] = self._image.flatten()[template_indices] else: image_batch[batch_indices] = self._temp_baseline[template_indices] # Reshape the image to proper sizes as required by the network image_batch = image_batch.reshape( -1, self.color_channels, self.width, self.height ) if index == self.__len__() - 1: # Only in the last run: Add the original image image_batch[batch_size - 1] = self._image else: # Save last image for the next run self._temp_image = image_batch[-1] # TODO return image_batch
5,848
35.55625
118
py
ExplainableAIImageMeasures
ExplainableAIImageMeasures-main/explainable_ai_image_measures/irof.py
import torch import numpy as np from skimage.segmentation import slic from explainable_ai_image_measures.pixel_manipulation import PixelManipulationBase class IrofDataset(PixelManipulationBase): def __init__( self, image, attribution, batch_size, irof_segments, irof_sigma, device, baseline_color ): PixelManipulationBase.__init__( self, image, attribution, False, batch_size, device, baseline_color ) self._irof_segments = irof_segments self._irof_sigma = irof_sigma self.generate_pixel_batches() self.generate_initial_image() self.generate_temp_baseline() def generate_pixel_batches(self): # Apply Slic algorithm to get superpixel areas img_np = self._image.detach().cpu().numpy().transpose(1, 2, 0) segments = slic(img_np, self._irof_segments, self._irof_segments).reshape(-1) nr_segments = np.max(segments) + 1 segments = torch.LongTensor(segments).to(device=self._device) # Attribution score of each segment = Mean attribution attr = self._attribution.reshape(-1) seg_mean = [ torch.mean(attr[segments == seg]).item() for seg in range(nr_segments) ] seg_mean = torch.FloatTensor(seg_mean).to(device=self._device) # Sort segments descending by mean attribution seg_rank = torch.argsort(-seg_mean) # Create lists of "shape" [batch_size, segment_size] # containing the indices of the segments self._pixel_batches = [[]] for seg in seg_rank: indices = torch.nonzero(segments == seg).flatten() IrofDataset._add_to_hierarhical_list( self._pixel_batches, self._batch_size, indices ) # Add placeholder for original image IrofDataset._add_to_hierarhical_list( self._pixel_batches, self._batch_size, torch.Tensor([0]).to(device=self._device) ) @staticmethod def _add_to_hierarhical_list(list_element, target_size, item): if len(list_element[-1]) == target_size: list_element.append([]) list_element[-1].append(item) def _gen_indices(self, index): batch_size = self._get_batch_size(index) # Get all pixels all_pixels = torch.cat(self._pixel_batches[index]).to(self._device).long() # Create a matrix of indices of size [batch_size, all_pixels] template_indices = all_pixels.view(1, -1).repeat(batch_size, 1) # For each package only keep the previous pixels and package size additional pixels pixel_per_image = torch.LongTensor( [len(package) for package in self._pixel_batches[index]] ).to(self._device) cumsum = torch.cumsum(pixel_per_image, dim=0) keep_index_template = torch.cat( [torch.arange(0, s.item()) for s in cumsum] ).to(self._device) template_indices = template_indices.reshape(-1)[keep_index_template] template_indices = self._color_channel_shift(template_indices) batch_indices = self._batch_shift(template_indices, pixel_per_image) return template_indices, batch_indices def _get_fake_image_size(self): return 1 def postprocess_scores(self, y): return 1-y
3,308
35.766667
95
py
ExplainableAIImageMeasures
ExplainableAIImageMeasures-main/explainable_ai_image_measures/pixel_relevancy.py
import torch from explainable_ai_image_measures.pixel_manipulation import PixelManipulationBase class PixelRelevancyDataset(PixelManipulationBase): def __init__(self, image, attribution, insert, batch_size, package_size, device, baseline_color): PixelManipulationBase.__init__( self, image, attribution, insert, batch_size, device, baseline_color ) self._package_size = package_size self._device = device self.generate_pixel_batches() self.generate_initial_image() self.generate_temp_baseline() def generate_pixel_batches(self): # For simplicity: Ensure that all packages have the same size. max_nr_pixels = ( self.width * self.height - self.width * self.height % self._package_size ) # Sort pixels in descending order by attribution score pixel_relevancy_desc = torch.argsort(-self._attribution.flatten())[ :max_nr_pixels ] # Add placeholder for original image placeholder = torch.LongTensor(self._package_size * [0]).to(device=self._device) pixel_relevancy_desc = torch.cat((pixel_relevancy_desc, placeholder)) # Form groups of size package_size pixel_relevancy_groups = pixel_relevancy_desc.reshape(-1, self._package_size) # Forms batches of groups: (batch_size x package_size) self._pixel_batches = torch.split( pixel_relevancy_groups, self._batch_size, dim=0 ) def _gen_indices(self, index): batch_size = self._get_batch_size(index) # Create a matrix of indices of size [batch_size, package_size * batch_size] template_indices = self._pixel_batches[index].view(1, -1).repeat(batch_size, 1) # For each package only keep the previous pixels and package_size additional pixels keep_index_template = torch.cat( [torch.arange(0, self._package_size * i) for i in range(1, batch_size + 1)] ) template_indices = template_indices.reshape(-1)[keep_index_template] template_indices = self._color_channel_shift(template_indices) # Shift each batch item by all the pixels of previous images pixel_per_image = torch.Tensor(batch_size * [self._package_size]).to(self._device).long() batch_indices = self._batch_shift(template_indices, pixel_per_image) return template_indices, batch_indices def _get_fake_image_size(self): return self._package_size def postprocess_scores(self, y): return y
2,552
41.55
101
py
ExplainableAIImageMeasures
ExplainableAIImageMeasures-main/explainable_ai_image_measures/__init__.py
from explainable_ai_image_measures.scoring_metric import Measures __version__ = "1.0.1" __all__ = ["Measures"]
112
21.6
65
py
ellip-corr
ellip-corr-master/PyEllipCorr.py
import os from numpy import f2py import numpy as np from collections import defaultdict from pyellip import select_phase, coeffs, ellip class PyEllipCorr: def __init__(self): self._tbl_fn = os.path.join(os.path.dirname(__file__), 'ellip/elcordir.tbl') self._coeffs = defaultdict(list) # end func def get_correction(self, phase, edist, edepth, ecolat, azim): """ :param phase: a string specifying the PHASE, -e.g P, ScP etc. :param edist: epicentral distance to station (degrees) :param edepth: depth of event (km) :param ecolat: epicentral co-latitude of source (degrees) :param azim: azimuth from source to station (degrees) :return: ellipticity-correction (s) """ ip, abrt = select_phase(phase, edist) if(abrt): print('Warning: Phase {} not found. Returning a null correction..'.format(phase)) return 0 # end if if ((phase, ip) not in self._coeffs.keys()): t0, t1, t2, d1,d2,delta, abrt = coeffs(self._tbl_fn, ip) if(not abrt): self._coeffs[(phase, ip)] = [t0, t1, t2, d1, d2, delta] return self.get_correction(phase, edist, edepth, ecolat, azim) else: return 0 # end if else: ecolat_rad = np.radians(ecolat) azim_rad = np.radians(azim) t0, t1, t2, d1, d2, delta = self._coeffs[(phase, ip)] tcor, abrt = ellip(t0, t1, t2, d1, d2, delta, edist, edepth, ecolat_rad, azim_rad); if(not abrt): return tcor else: return 0 # end if # end func # end class
1,696
32.94
95
py
ellip-corr
ellip-corr-master/__init__.py
0
0
0
py
ellip-corr
ellip-corr-master/tests/test_ellipticity_corr.py
from __future__ import print_function import pytest from PyEllipCorr import PyEllipCorr """ The values below were extracted from ttimel """ """ Source latitude: -30 Source depth (km): 124 Azimuth from source: 39 delta: 65 # code time time(el) dT/dD dT/dh d2T/dD2 """ data= """ 1 P 626.6652 626.8218 6.4671 -1.09E-01 -4.26E-03 2 pP 656.8627 657.0470 6.5478 1.09E-01 -4.21E-03 3 PcP 658.6354 658.7878 4.1323 -1.18E-01 1.38E-02 4 sP 670.3326 670.4863 6.5283 2.14E-01 -4.23E-03 5 PP 771.0760 771.6174 8.7483 -9.46E-02 -1.25E-02 6 PKiKP 1023.1398 1023.5211 1.3232 -1.23E-01 1.85E-02 7 pPKiKP 1056.6406 1057.0413 1.3198 1.23E-01 1.85E-02 8 sPKiKP 1069.3484 1069.7469 1.3205 2.22E-01 1.85E-02 9 S 1138.7920 1139.0929 12.2313 -1.92E-01 -2.71E-03 10 SPn 1156.3033 1156.9156 13.6603 -1.83E-01 -2.31E-02 11 pS 1173.5881 1173.9027 12.5499 4.63E-02 -2.35E-03 12 PnS 1176.5562 1177.1404 13.3906 -1.75E-02 -5.56E-03 13 sS 1190.7726 1191.1244 12.3694 1.91E-01 -2.67E-03 14 SKSac 1209.6849 1209.9666 7.5880 -2.11E-01 -1.13E-01 15 SKKSac 1209.6868 1209.9685 7.5903 -2.11E-01 -4.81E-01 16 ScS 1209.8174 1210.1000 7.6990 -2.11E-01 7.04E-03 17 SKiKP 1224.7056 1225.1292 1.3785 -2.22E-01 1.83E-02 18 pSKSac 1252.3256 1252.6616 7.5901 1.03E-01 -2.07E-01 19 sSKSac 1266.1160 1266.4423 7.5896 2.11E-01 -1.63E-01 20 SS 1392.8646 1393.8425 15.5487 -1.70E-01 -9.15E-03 21 PKKPdf 1850.6364 1853.4150 -1.3739 -1.23E-01 -1.69E-02 22 SKKPdf 2052.1799 2055.6609 -1.3144 -2.22E-01 -1.71E-02 23 PKKSdf 2064.8875 2068.9170 -1.3136 -1.23E-01 -1.71E-02 24 SKKSdf 2266.2419 2271.1333 -1.2580 -2.22E-01 -1.75E-02 25 P'P'df 2350.1360 2348.5884 -1.6599 -1.23E-01 -2.16E-02 26 P'P'bc 2355.4346 2353.8870 -2.7297 -1.21E-01 -3.67E-03 27 P'P'ab 2359.7952 2358.2476 -3.9679 -1.19E-01 7.92E-03 28 S'S'df 3197.0603 3197.8882 -1.3915 -2.22E-01 -1.70E-02""" data = data.split() ellipcorr_dict = {} pyellip = PyEllipCorr() for i, d in enumerate(data): if i % 7 == 1: k = d if i % 7 == 2: tt = float(d) if i % 7 == 3: tcor = float(d) - tt ellipcorr_dict[k] = tcor @pytest.fixture(params=list(ellipcorr_dict.keys())) def tcor_pair(request): return request.param, ellipcorr_dict[request.param] def test_ellipticity_corr(tcor_pair): phase, corr = tcor_pair result = pyellip.get_correction(phase, 65, 124, (90 - -30), 39) print(result, corr) assert abs(result - corr) < 1.0e-3
3,076
42.957143
80
py
ellip-corr
ellip-corr-master/ellip/__init__.py
0
0
0
py
ellip-corr
ellip-corr-master/tau/__init__.py
0
0
0
py
zpdgen
zpdgen-master/itg_ai_loc.py
import numpy as np import matplotlib.pyplot as plt import gpdf as gp from scipy.optimize import root etai=2.5 LnbyR=0.2 rbyR=0.18 kpar=0.1 tau=1.0 def epsfun(v): om=v[0]+1j*v[1] omsi=-ky omdi=2*omsi*LnbyR za=-om/omdi zb=-np.sqrt(2)*kpar/omdi b=ky**2 i10=gp.Inm(za,zb,b,1,0) i12=gp.Inm(za,zb,b,1,2) i30=gp.Inm(za,zb,b,3,0) eps=1+1/tau+(i10*(om-omsi*(1-1.5*etai))-omsi*etai*(i12+i30))/omdi res=[float(np.real(eps)),float(np.imag(eps))] return res kys=np.arange(0.01,3.0,0.01) ky0=0.7 om0=[-0.2,0.2] omky=np.zeros(np.shape(kys))*(1+1j) ind0=np.argmin((kys-ky0)**2) numk=len(kys) inds=np.concatenate((np.arange(ind0,numk), np.arange(ind0-1,-1,-1))) for l in inds: ky=kys[l] res=root(epsfun,om0,tol=1e-8,method='hybr') omky[l]=res.x[0]+1j*res.x[1] if(l==numk-1): om0=[float(np.real(omky[ind0])),float(np.imag(omky[ind0]))] else: om0=res.x plt.plot(kys,np.imag(omky),'x-') #plt.axis([0.0,2.0,-0.1,0.2]) #plt.plot(kys,-np.real(omky),'r') plt.show()
1,029
21.888889
69
py
zpdgen
zpdgen-master/itg_ai_loc_mat.py
import numpy as np import matplotlib.pyplot as plt import gpdf as gp from scipy.optimize import root etai=2.5 LnbyR=0.2 rbyR=0.18 kpar=0.1 tau=1.0 def epsfun(v): om=v[0]+1j*v[1] omsi=-ky omdi=2*omsi*LnbyR za=-om/omdi zb=-np.sqrt(2)*kpar/omdi b=ky**2 anm=np.zeros((4,3),dtype=np.complex128) anm[1,0]=(om-omsi*(1-1.5*etai))/omdi anm[1,2]=-omsi*etai/omdi anm[3,0]=-omsi*etai/omdi eps=1+1/tau+gp.sigmazpd(za,zb,b,anm) res=[float(np.real(eps)),float(np.imag(eps))] return res kys=np.arange(0.01,3.0,0.01) ky0=0.7 om0=[-0.2,0.2] omky=np.zeros(np.shape(kys))*(1+1j) ind0=np.argmin((kys-ky0)**2) numk=len(kys) inds=np.concatenate((np.arange(ind0,numk), np.arange(ind0-1,-1,-1))) for l in inds: ky=kys[l] # print('ky:',ky) res=root(epsfun,om0,tol=1e-8,method='hybr') omky[l]=res.x[0]+1j*res.x[1] if(l==numk-1): om0=[float(np.real(omky[ind0])),float(np.imag(omky[ind0]))] else: om0=res.x plt.plot(kys,np.imag(omky),'x-') #plt.axis([0.0,2.0,-0.1,0.2]) #plt.plot(kys,-np.real(omky),'r') plt.show()
1,084
22.085106
68
py
zpdgen
zpdgen-master/plot_eps.py
import numpy as np import matplotlib.pyplot as plt import gpdf as gp etai=2.5 LnbyR=0.2 rbyR=0.18 kpar=0.1 tau=1.0 ky=0.06 def epsfun(v): om=v omsi=-ky omdi=2*omsi*LnbyR za=-om/omdi zb=-kpar/omdi*np.sqrt(2) b=ky**2 i10=gp.Inm(za,zb,b,1,0) i12=gp.Inm(za,zb,b,1,2) i30=gp.Inm(za,zb,b,3,0) eps=1+1/tau+(i10*(om-omsi*(1-1.5*etai))-omsi*etai*(i12+i30))/omdi return eps xx,yy=np.meshgrid(np.arange(-1,1,0.01),np.arange(-1,1,0.01)) om=xx+1j*yy; cnts=np.arange(-2,2.2,0.2) wdts=np.ones(cnts.shape); wdts[0]=2.0 wdts[5]=2.0 wdts[10]=4.0 wdts[15]=2.0 wdts[20]=2.0 eps=epsfun(om) #gp.Inm(om,0.0,0.09,1,0) plt.pcolormesh(np.real(om),np.imag(om),np.imag(eps),shading='gouraud',rasterized=True) plt.clim(-0.01,0.01) plt.colorbar() plt.contour(np.real(om),np.imag(om),np.real(eps),[0.0],colors='k') plt.contour(np.real(om),np.imag(om),np.imag(eps),[0.0],colors='r') plt.show()
908
21.170732
86
py
zpdgen
zpdgen-master/py_time.py
import numpy as np import matplotlib.pyplot as plt import gpdf as gp import time nlist=[[1,0],[1,2],[3,0]] ii=0; xx,yy=np.meshgrid(np.arange(-6,6,0.1),np.arange(-6,6,0.1)) za=xx+1j*yy; zb=0.0 b=0.09 for ns in nlist: ii=ii+1; [n,m]=ns; print('computing I'+str(n)+str(m)+' ...') t0 = time.clock() inm=gp.Inm(za,zb,b,n,m) print(time.clock()-t0,"seconds")
376
19.944444
58
py
zpdgen
zpdgen-master/gpdf.py
from numpy import reshape,shape,rank,transpose from inmzpd import inmweid,gmweid from epszpd import epsweid,sigweid def Inm(za,zb,b,n,m): res=inmweid(za,zb,b,n,m) if (shape(za) ==()): res=res[0] return res res=reshape(res,shape(transpose(za))) res=transpose(res) return res def epsitg(om,pars): res=epsweid(om,pars) if (shape(om) ==()): res=res[0] return res res=reshape(res,shape(transpose(om))) res=transpose(res) return res def sigmazpd(za,zb,b,anm): if (rank(anm)==2): numn,numm=shape(anm) elif (rank(anm)==1): numn=size(anm) numm=0 elif (rank(anm)==0): numn=0 numm=0 else: print("error the rank of anm is not 0,1 or 2") return 0 res=sigweid(za,zb,b,anm,numn,numm) if (shape(za) ==()): res=res[0] return res res=reshape(res,shape(transpose(za))) res=transpose(res) return res def Gm(z1,z2,m): if(shape(z1)!=shape(z2)): print("error: shape(z1)!=shape(z2)!\n"); res=() return res; res=gmweid(z1,z2,m) if (shape(z1) ==() and shape(z2)==()): res=res[0] return res res=reshape(res,shape(transpose(z1))) res=transpose(res) return res
1,280
22.290909
54
py
zpdgen
zpdgen-master/plot_inm_re.py
import numpy as np import matplotlib.pyplot as plt import gpdf as gp nlist=[[1,0],[2,0],[3,0],[1,2],[2,2],[3,2],[1,4],[2,4],[3,4]] cnts=np.arange(-2,2.2,0.2) wdts=np.ones(cnts.shape); wdts[0]=2.0 wdts[5]=2.0 wdts[10]=4.0 wdts[15]=2.0 wdts[20]=2.0 vcb = np.arange(-12,14,2) ii=0; xx,yy=np.meshgrid(np.arange(-6,6,0.1),np.arange(-6,6,0.1)) za=xx+1j*yy; zb=0.0 b=0.09 for ns in nlist: ii=ii+1; [n,m]=ns; print('computing I'+str(n)+str(m)+' ...\n') inm=gp.Inm(za,zb,b,n,m) plt.subplot(3,3,ii) plt.pcolormesh(np.real(za),np.imag(za),np.real(inm),shading='gouraud',rasterized=True) plt.clim(-10,10) cs=plt.contour(np.real(za),np.imag(za),np.real(inm),cnts,colors='k',linewidths=wdts*0.5) aa=r'$Re[I_{'+str(n)+str(m)+r'}(\zeta_\alpha,\zeta_\beta,b)]$' plt.text(-1,3,aa,fontsize=14) if (m!=4 and n!=1): plt.tick_params(labelbottom='off',labelleft='off') elif(n==1 and m==4) : plt.text(2,-7.5,r'$Re[\zeta_\alpha]$',fontsize=14) plt.text(-7.5,3,r'$Im[\zeta_\alpha]$',fontsize=14,rotation=90) plt.tick_params(axis='both', which='major', labelsize=9,labelbottom='on',labelleft='on') elif(m==4) : plt.text(2,-7.5,r'$Re[\zeta_\alpha]$',fontsize=14) plt.tick_params(axis='both', which='major', labelsize=9,labelbottom='on',labelleft='off') elif(n==1): plt.text(-7.5,3,r'$Im[\zeta_\alpha]$',fontsize=14,rotation=90) plt.tick_params(axis='both', which='major', labelsize=9,labelleft='on',labelbottom='off') plt.subplots_adjust(wspace=0.02,hspace=0.02,left=0.05,right=0.92,top=0.95,bottom=0.05) fig=plt.gcf() cax = fig.add_axes([0.93, 0.1, 0.02, 0.80]) plt.figure(2) s=plt.contourf(np.real(za),np.imag(za),np.real(inm)*(np.real(inm)<14)*(np.real(inm)>-14),100) plt.clim(-10,10) plt.close() plt.figure(1) cbar=plt.colorbar(s, cax=cax,ticks=vcb) cbar.ax.tick_params(labelsize=8) cbar.solids.set_rasterized(True) fig.set_size_inches(8,8) plt.savefig("figures_re.pdf")
1,970
35.5
97
py
zpdgen
zpdgen-master/plot_itg_jykim94.py
import numpy as np import matplotlib.pyplot as plt import gpdf as gp from scipy.optimize import root etai=2.5 LnbyR=0.2 rbyR=0.18 kpar=0.1 tau=1.0 def epsfun(v): om=v[0]+1j*v[1] omsi=-ky omdi=2*omsi*LnbyR za=-om/omdi zb=-np.sqrt(2)*kpar/omdi b=ky**2 i10=gp.Inm(za,zb,b,1,0) i12=gp.Inm(za,zb,b,1,2) i30=gp.Inm(za,zb,b,3,0) eps=1+1/tau+(i10*(om-omsi*(1-1.5*etai))-omsi*etai*(i12+i30))/omdi res=[float(np.real(eps)),float(np.imag(eps))] return res kys=np.arange(0.01,2.4,0.01) ky0=0.7 om0=[-0.2,0.2] omky=np.zeros(np.shape(kys))*(1+1j) ind0=np.argmin((kys-ky0)**2) numk=len(kys) inds=np.concatenate((np.arange(ind0,numk), np.arange(ind0-1,-1,-1))) for l in inds: ky=kys[l] res=root(epsfun,om0,tol=1e-8,method='hybr') omky[l]=res.x[0]+1j*res.x[1] if(l==numk-1): om0=[float(np.real(omky[ind0])),float(np.imag(omky[ind0]))] else: om0=res.x phi=np.loadtxt('jykim94a.dat') ax1=plt.subplot(1,2,1) plt.plot(kys,np.imag(omky),color='k',linewidth=2) plt.plot(phi[:,0],phi[:,1],'--',color='r') plt.xlabel('$k_y$',fontsize=18) plt.ylabel('$\gamma$',fontsize=18) plt.tight_layout() #plt.axis([0.0,2.0,-0.1,0.2]) #plt.plot(kys,-np.real(omky),'r') #plt.show() x0,x1 = ax1.get_xlim() y0,y1 = ax1.get_ylim() ax1.set_aspect((x1-x0)/(y1-y0)) ax1=plt.subplot(1,2,2) phi=np.loadtxt('jykim94a_re.dat') plt.plot(kys,np.real(omky),color='k',linewidth=2) plt.plot(phi[:,0],-phi[:,1],'--',color='r') plt.xlabel('$k_y$',fontsize=18) plt.ylabel('$\omega_r$',fontsize=18) plt.tight_layout() x0,x1 = ax1.get_xlim() y0,y1 = ax1.get_ylim() ax1.set_aspect((x1-x0)/(y1-y0)) plt.savefig("fig_kim94a.pdf") gam=np.real(omky) np.savetxt('test.txt',gam, delimiter=" ")
1,715
24.61194
69
py
zpdgen
zpdgen-master/itg_ai_loc_comb.py
import numpy as np import matplotlib.pyplot as plt import gpdf as gp from scipy.optimize import root etai=2.5 LnbyR=0.2 rbyR=0.18 kpar=0.1 tau=1.0 def epsfun(v): om=v[0]+1j*v[1] omsi=-ky omdi=2*omsi*LnbyR za=-om/omdi zb=-np.sqrt(2)*kpar/omdi b=ky**2 pars=(omdi,etai,tau,ky,kpar) eps=gp.epsitg(om,pars) res=[float(np.real(eps)),float(np.imag(eps))] return res kys=np.arange(0.01,3.0,0.01) ky0=0.7 om0=[-0.2,0.2] omky=np.zeros(np.shape(kys))*(1+1j) ind0=np.argmin((kys-ky0)**2) numk=len(kys) inds=np.concatenate((np.arange(ind0,numk), np.arange(ind0-1,-1,-1))) for l in inds: ky=kys[l] res=root(epsfun,om0,tol=1e-8,method='hybr') omky[l]=res.x[0]+1j*res.x[1] if(l==numk-1): om0=[float(np.real(omky[ind0])),float(np.imag(omky[ind0]))] else: om0=res.x plt.plot(kys,np.imag(omky),'x-') #plt.axis([0.0,2.0,-0.1,0.2]) #plt.plot(kys,-np.real(omky),'r') plt.show()
935
20.767442
68
py
zpdgen
zpdgen-master/plot_inm_im.py
import numpy as np import matplotlib.pyplot as plt import gpdf as gp nlist=[[1,0],[2,0],[3,0],[1,2],[2,2],[3,2],[1,4],[2,4],[3,4]] cnts=np.arange(-2,2.2,0.2) wdts=np.ones(cnts.shape); wdts[0]=2.0 wdts[5]=2.0 wdts[10]=4.0 wdts[15]=2.0 wdts[20]=2.0 vcb = np.arange(-12,14,2) ii=0; xx,yy=np.meshgrid(np.arange(-6,6,0.1),np.arange(-6,6,0.1)) za=xx+1j*yy; zb=0.0 b=0.09 for ns in nlist: ii=ii+1; [n,m]=ns; print('computing I'+str(n)+str(m)+' ...\n') inm=gp.Inm(za,zb,b,n,m) plt.subplot(3,3,ii) plt.pcolormesh(np.real(za),np.imag(za),np.imag(inm),shading='gouraud',rasterized=True) plt.clim(-10,10) cs=plt.contour(np.real(za),np.imag(za),np.imag(inm),cnts,colors='k',linewidths=wdts*0.5) aa=r'$Im[I_{'+str(n)+str(m)+r'}(\zeta_\alpha,\zeta_\beta,b)]$' plt.text(-1,3,aa,fontsize=14) if (m!=4 and n!=1): plt.tick_params(labelbottom='off',labelleft='off') elif(n==1 and m==4) : plt.text(2,-7.5,r'$Re[\zeta_\alpha]$',fontsize=14) plt.text(-7.5,3,r'$Im[\zeta_\alpha]$',fontsize=14,rotation=90) plt.tick_params(axis='both', which='major', labelsize=9,labelbottom='on',labelleft='on') elif(m==4) : plt.text(2,-7.5,r'$Re[\zeta_\alpha]$',fontsize=14) plt.tick_params(axis='both', which='major', labelsize=9,labelbottom='on',labelleft='off') elif(n==1): plt.text(-7.5,3,r'$Im[\zeta_\alpha]$',fontsize=14,rotation=90) plt.tick_params(axis='both', which='major', labelsize=9,labelleft='on',labelbottom='off') plt.subplots_adjust(wspace=0.02,hspace=0.02,left=0.05,right=0.92,top=0.95,bottom=0.05) fig=plt.gcf() cax = fig.add_axes([0.93, 0.1, 0.02, 0.80]) plt.figure(2) s=plt.contourf(np.real(za),np.imag(za),np.imag(inm)*(np.imag(inm)<14)*(np.imag(inm)>-14),100) plt.clim(-10,10) plt.close() plt.figure(1) cbar=plt.colorbar(s, cax=cax,ticks=vcb) cbar.ax.tick_params(labelsize=8) cbar.solids.set_rasterized(True) fig.set_size_inches(8,8) plt.savefig("figures_im.pdf")
1,970
35.5
97
py
GAStimator
GAStimator-master/setup.py
from setuptools import setup with open("README.md", "r") as fh: long_description = fh.read() setup(name='gastimator', version='0.4.7', description='Implementation of a Python MCMC gibbs-sampler with adaptive stepping', url='https://github.com/TimothyADavis/GAStimator', author='Timothy A. Davis', author_email='DavisT@cardiff.ac.uk', long_description=long_description, long_description_content_type="text/markdown", license='GNU GPLv3', packages=['gastimator'], install_requires=[ 'numpy', 'matplotlib', 'tqdm', 'plotbin', 'joblib>=0.16.0', ], classifiers=[ 'Development Status :: 3 - Alpha', 'License :: OSI Approved :: GNU General Public License v3 (GPLv3)', 'Programming Language :: Python :: 3', 'Operating System :: OS Independent', ], zip_safe=False)
962
29.09375
90
py
GAStimator
GAStimator-master/gastimator/priors.py
#!/usr/bin/env python3 # coding: utf-8 import numpy as np class priors: def __init__(): pass class gaussian: def __init__(self,mu,sigma): self.mu=mu self.sigma=sigma def eval(self,x,**kwargs): xs = (x - self.mu) / self.sigma return (-(xs*xs)/2.0) - np.log(2.5066282746310002*self.sigma) #returns ln(prior)
422
22.5
92
py
GAStimator
GAStimator-master/gastimator/__init__.py
#!/usr/bin/env python3 # coding: utf-8 from gastimator.gastimator import gastimator from gastimator.priors import priors
120
29.25
44
py
GAStimator
GAStimator-master/gastimator/corner_plot.py
############################################################################## # # This is the history of Michele Cappellari modifications # to Daniel Foreman-Mackey corner_plot routine. # # V1.0.0: Included "like" and "xstry" optional inputs # to show individual points coloured by their likelihood # and to show all points tried by MCMC. # Michele Cappellari, Oxford, 12 January 2014 # V1.1.0: Added "init" keyword to simply initialize an empty window. # MC, Oxford, 10 October 2014 # V1.1.1: Check that input sizes of "extents" and "labels" match. # MC, Portsmouth, 15 October 2014 # V1.1.2: Only show unique dots in "xs" and "like", but properly include # duplicates in histograms. MC, Oxford, 4 November 2014 # V1.1.3: Allow for scaling of axes labels. # Fix program stop when values are the same within numerical accuracy # but np.std(x) is not exactly zero. MC, Oxford, 10 December 2014 # V1.1.4: Fix potential program stop with few input values. # MC, Oxford, 20 November 2015 # V1.1.5: Included histo_bin_width(). MC, Oxford, 8 February 2017 # V1.1.6: Added `rasterized` keyword and default value based on points number. # Use plt.plot for try values. MC, Oxford, 25 January 2018 # V1.1.7: Changes definition of `extents` for consistency with Scipy's `bounds`. # MC, Oxford, 27 April 2018 # ############################################################################## from __future__ import print_function, absolute_import, unicode_literals __all__ = ["corner_plot", "hist2d", "error_ellipse"] __version__ = "0.0.6" __author__ = "Dan Foreman-Mackey (danfm@nyu.edu)" __copyright__ = "Copyright 2013 Daniel Foreman-Mackey" __contributors__ = [ # Alphabetical by first name. "Adrian Price-Whelan @adrn", "Brendon Brewer @eggplantbren", "Ekta Patel @ekta1224", "Emily Rice @emilurice", "Geoff Ryan @geoffryan", "Kyle Barbary @kbarbary", "Phil Marshall @drphilmarshall", "Pierre Gratier @pirg", ] import numpy as np import matplotlib.pyplot as plt from matplotlib.ticker import MaxNLocator from matplotlib.colors import LinearSegmentedColormap from matplotlib.patches import Ellipse import matplotlib.cm as cm from plotbin.sauron_colormap import register_sauron_colormap ###################################################################### def corner_plot(xs, like=None, xstry=None, weights=None, labels=None, extents=None, truths=None, truth_color="steelblue", fignum=None, scale_hist=False, quantiles=[], verbose=True, init=False, plot_contours=True, plot_datapoints=True, fig=None, labels_scaling=1, rasterized=None, **kwargs): """ Make a *sick* corner plot showing the projections of a data set in a multi-dimensional space. kwargs are passed to hist2d() or used for `matplotlib` styling. Parameters ---------- xs : array_like (nsamples, ndim) The samples. This should be a 1- or 2-dimensional array. For a 1-D array this results in a simple histogram. For a 2-D array, the zeroth axis is the list of samples and the next axis are the dimensions of the space. xstry : array_like (nsamples, ndim) (optional) Contains all tried parameter, not only the accepted moves. like : array_like (nsamples) (optional) Likelihood of each xs[j, :] set of parameters to be shown on the plot. weights : array_like (nsamples,) The weight of each sample. If `None` (default), samples are given equal weight. labels : iterable (ndim,) (optional) A list of names for the dimensions. extents : 2-tuple of array_like (2, ndim) Lower and upper bounds on independent variables. Each array must match the size ndim, e.g., [(0., 10., 30., 25.), (1., 15, 40., 55.)]. truths : iterable (ndim,) (optional) A list of reference values to indicate on the plots. truth_color : str (optional) A ``matplotlib`` style color for the ``truths`` makers. scale_hist : bool (optional) Should the 1-D histograms be scaled in such a way that the zero line is visible? quantiles : iterable (optional) A list of fractional quantiles to show on the 1-D histograms as vertical dashed lines. verbose : bool (optional) If true, print the values of the computed quantiles. plot_contours : bool (optional) Draw contours for dense regions of the plot. plot_datapoints : bool (optional) Draw the individual data points. fig : matplotlib.Figure (optional) Overplot onto the provided figure object. """ # Deal with 1D sample lists. xs = np.atleast_1d(xs) if len(xs.shape) == 1: xs = np.atleast_2d(xs) else: assert len(xs.shape) == 2, "The input sample array must be 1- or 2-D." xs = xs.T if xstry is not None: xstry = xstry.T if labels is not None: # note xs was already trasposed assert len(labels) == xs.shape[0], 'lengths of labels must match number of dimensions' if extents is not None: extents = np.array(extents).T assert len(extents) == xs.shape[0], 'lengths of extents must match number of dimensions' if weights is not None: weights = np.asarray(weights) assert weights.ndim == 1, 'weights must be 1-D' assert xs.shape[1] == weights.shape[0], 'lengths of weights must match number of samples' if "sauron" not in cm._colormaps.keys(): register_sauron_colormap() # backwards-compatibility plot_contours = kwargs.get("smooth", plot_contours) K = len(xs) if fig is None: factor = 2.0 # size of one side of one panel lbdim = 0.5 * factor # size of left/bottom margin trdim = 0.05 * factor # size of top/right margin whspace = 0.05 # w/hspace size plotdim = factor * K + factor * (K - 1.) * whspace dim = lbdim + plotdim + trdim fig, axes = plt.subplots(K, K, figsize=(dim, dim), num=fignum) lb = lbdim / dim tr = (lbdim + plotdim) / dim fig.subplots_adjust(left=lb, bottom=lb, right=tr, top=tr, wspace=whspace, hspace=whspace) if init: return fig else: try: axes = np.array(fig.axes).reshape((K, K)) except: raise ValueError("Provided figure has {0} axes, but data has " "dimensions K={1}".format(len(fig.axes), K)) if extents is None: extents = [[x.min(), x.max()] for x in xs] # Check for parameters that never change. m = np.array([e[0] == e[1] for e in extents], dtype=bool) if np.any(m): raise ValueError(("It looks like the parameter(s) in column(s) " "{0} have no dynamic range. Please provide an " "`extent` argument.") .format(", ".join(map("{0}".format, np.arange(len(m))[m])))) else: # If any of the extents are percentiles, convert them to ranges. for i in range(len(extents)): try: emin, emax = extents[i] except TypeError: q = [0.5 - 0.5*extents[i], 0.5 + 0.5*extents[i]] extents[i] = quantile(xs[i], q, weights=weights) # Extract unique values to reduce the memory taken by the figure (MC) # The unique version is used for the scatter plots but not for the histograms if like is not None: likeUnique, w = np.unique(like, return_index=True) xsUnique = xs[:, w] w = np.argsort(likeUnique) # Sort to plot most likely values last (MC) mx = likeUnique[w[-1]] # Maximum likelihood value so far likeUnique = likeUnique[w] xsUnique = xsUnique[:, w] if rasterized is None: rasterized = True if xsUnique.size > 1e4 else False for i, x in enumerate(xs): ax = axes[i, i] # Plot the histograms. Scott's rule for bin size # For a Normal distribution sigma=1.4826*MAD, so the robust version of # Scott (1979) rule: binsize = 3.49*sigma/n^{1/3} is binsize = 5.17*MAD/n^{1/3} # binsize = 5.17*np.median(np.abs(x - np.median(x))) / x.size**(1./3.) binsize = 3.49*np.std(x) / x.size**(1./3.) # binsize = histo_bin_width(x) if binsize > 0: nbins = int((extents[i][1] - extents[i][0])/binsize) # if nbins > 20: # nbins //= 2 # divide by two for aesthetic reasons and keep integer if nbins > 1e4: # x values are the same within numerical accuracy nbins = 5 else: nbins = 5 ax.cla() # clean current subplot before plotting n, b, p = ax.hist(x, weights=weights, bins=kwargs.get("bins", nbins), range=extents[i], color=kwargs.get("color", "b"), histtype='stepfilled') # Axes labels larger than tick labels ax.yaxis.label.set_size(plt.rcParams['font.size']*labels_scaling) ax.xaxis.label.set_size(plt.rcParams['font.size']*labels_scaling) if truths is not None: ax.axvline(truths[i], color=truth_color) # Plot quantiles if wanted. if len(quantiles) > 0: qvalues = quantile(x, quantiles, weights=weights) for q in qvalues: ax.axvline(q, ls="dashed", color=kwargs.get("color", "k")) if verbose: print("Quantiles:") print([item for item in zip(quantiles, qvalues)]) # Set up the axes. ax.set_xlim(extents[i]) if scale_hist: maxn = np.max(n) ax.set_ylim(-0.1 * maxn, 1.1 * maxn) else: ax.set_ylim(0, 1.1 * np.max(n)) ax.set_yticklabels([]) ax.xaxis.set_major_locator(MaxNLocator(5)) # Not so DRY. if i < K - 1: ax.set_xticklabels([]) else: [l.set_rotation(45) for l in ax.get_xticklabels()] if labels is not None: ax.set_xlabel(labels[i]) ax.xaxis.set_label_coords(0.5, -0.3) for j, y in enumerate(xs): ax = axes[i, j] if j > i: ax.set_visible(False) ax.set_frame_on(False) continue elif j == i: continue ax.cla() # clean current subplot before plotting ax.yaxis.label.set_size(plt.rcParams['font.size']*labels_scaling) ax.xaxis.label.set_size(plt.rcParams['font.size']*labels_scaling) if xstry is not None: ax.plot(xstry[j,:], xstry[i,:], ',m', ms=1, rasterized=rasterized, zorder=0) # Plot black when DeltaLogLike=9/2-->DeltaChi2=9 (3sigma) if like is not None: ax.scatter(xsUnique[j,:], xsUnique[i,:], c=likeUnique, vmin=mx-4.5, vmax=mx, edgecolors='None', cmap='sauron', rasterized=rasterized, **kwargs) ax.set_xlim(extents[j]) ax.set_ylim(extents[i]) else: hist2d(y, x, ax=ax, extent=[extents[j], extents[i]], plot_contours=plot_contours, plot_datapoints=plot_datapoints, weights=weights, **kwargs) # ax.tick_params(width=2, which='major') if truths is not None: ax.plot(truths[j], truths[i], "s", color=truth_color) ax.axvline(truths[j], color=truth_color) ax.axhline(truths[i], color=truth_color) ax.xaxis.set_major_locator(MaxNLocator(5)) ax.yaxis.set_major_locator(MaxNLocator(5)) if i < K - 1: ax.set_xticklabels([]) else: [l.set_rotation(45) for l in ax.get_xticklabels()] if labels is not None: ax.set_xlabel(labels[j]) ax.xaxis.set_label_coords(0.5, -0.3) if j > 0: ax.set_yticklabels([]) else: [l.set_rotation(45) for l in ax.get_yticklabels()] if labels is not None: ax.set_ylabel(labels[i]) ax.yaxis.set_label_coords(-0.3, 0.5) return fig ###################################################################### def quantile(x, q, weights=None): """ Like numpy.percentile, but: * Values of q are quantiles [0., 1.] rather than percentiles [0., 100.] * scalar q not supported (q must be iterable) * optional weights on x """ if weights is None: return np.percentile(x, [100. * qi for qi in q]) else: idx = np.argsort(x) xsorted = x[idx] cdf = np.add.accumulate(weights[idx]) cdf /= cdf[-1] return np.interp(q, cdf, xsorted).tolist() ###################################################################### def error_ellipse(mu, cov, ax=None, factor=1.0, **kwargs): """ Plot the error ellipse at a point given its covariance matrix. """ # some sane defaults facecolor = kwargs.pop('facecolor', 'none') edgecolor = kwargs.pop('edgecolor', 'k') x, y = mu U, S, V = np.linalg.svd(cov) theta = np.degrees(np.arctan2(U[1, 0], U[0, 0])) ellipsePlot = Ellipse(xy=[x, y], width=2 * np.sqrt(S[0]) * factor, height=2 * np.sqrt(S[1]) * factor, angle=theta, facecolor=facecolor, edgecolor=edgecolor, **kwargs) if ax is None: ax = plt.gca() ax.add_patch(ellipsePlot) return ellipsePlot ###################################################################### def hist2d(x, y, *args, **kwargs): """ Plot a 2-D histogram of samples. """ ax = kwargs.pop("ax", plt.gca()) extent = kwargs.pop("extent", [[x.min(), x.max()], [y.min(), y.max()]]) bins = kwargs.pop("bins", 50) color = kwargs.pop("color", "k") linewidths = kwargs.pop("linewidths", None) plot_datapoints = kwargs.get("plot_datapoints", True) plot_contours = kwargs.get("plot_contours", True) cmap = cm.get_cmap("gray") cmap._init() cmap._lut[:-3, :-1] = 0. cmap._lut[:-3, -1] = np.linspace(1, 0, cmap.N) X = np.linspace(extent[0][0], extent[0][1], bins + 1) Y = np.linspace(extent[1][0], extent[1][1], bins + 1) try: H, X, Y = np.histogram2d(x.flatten(), y.flatten(), bins=(X, Y), weights=kwargs.get('weights', None)) except ValueError: raise ValueError("It looks like at least one of your sample columns " "have no dynamic range. You could try using the " "`extent` argument.") V = 1.0 - np.exp(-0.5 * np.arange(0.5, 2.1, 0.5) ** 2) Hflat = H.flatten() inds = np.argsort(Hflat)[::-1] Hflat = Hflat[inds] sm = np.cumsum(Hflat) sm /= sm[-1] for i, v0 in enumerate(V): try: V[i] = Hflat[sm <= v0][-1] except: V[i] = Hflat[0] X1, Y1 = 0.5 * (X[1:] + X[:-1]), 0.5 * (Y[1:] + Y[:-1]) X, Y = X[:-1], Y[:-1] if plot_datapoints: ax.plot(x, y, "o", color=color, ms=1.5, zorder=-1, alpha=0.1, rasterized=True) if plot_contours: ax.contourf(X1, Y1, H.T, [V[-1], H.max()], cmap=LinearSegmentedColormap.from_list("cmap", ([1] * 3, [1] * 3), N=2), antialiased=False) if plot_contours: ax.pcolor(X, Y, H.max() - H.T, cmap=cmap) ax.contour(X1, Y1, H.T, V, colors=color, linewidths=linewidths) data = np.vstack([x, y]) mu = np.mean(data, axis=1) cov = np.cov(data) if kwargs.pop("plot_ellipse", False): error_ellipse(mu, cov, ax=ax, edgecolor="r", ls="dashed") ax.set_xlim(extent[0]) ax.set_ylim(extent[1]) ###################################################################### def _hist_bin_sturges(x): """ Sturges histogram bin estimator. A very simplistic estimator based on the assumption of normality of the data. This estimator has poor performance for non-normal data, which becomes especially obvious for large data sets. The estimate depends only on size of the data. Parameters ---------- x : array_like Input data that is to be histogrammed, trimmed to range. May not be empty. Returns ------- h : An estimate of the optimal bin width for the given data. """ return x.ptp() / (np.log2(x.size) + 1.0) ###################################################################### def _hist_bin_fd(x): """ The Freedman-Diaconis histogram bin estimator. The Freedman-Diaconis rule uses interquartile range (IQR) to estimate binwidth. It is considered a variation of the Scott rule with more robustness as the IQR is less affected by outliers than the standard deviation. However, the IQR depends on fewer points than the standard deviation, so it is less accurate, especially for long tailed distributions. If the IQR is 0, this function returns 1 for the number of bins. Binwidth is inversely proportional to the cube root of data size (asymptotically optimal). Parameters ---------- x : array_like Input data that is to be histogrammed, trimmed to range. May not be empty. Returns ------- h : An estimate of the optimal bin width for the given data. """ iqr = np.subtract(*np.percentile(x, [75, 25])) return 2.0 * iqr * x.size ** (-1.0 / 3.0) ###################################################################### def histo_bin_width(x): """ This function is copied from Numpy 1.11 Histogram bin estimator that uses the minimum width of the Freedman-Diaconis and Sturges estimators. The FD estimator is usually the most robust method, but its width estimate tends to be too large for small `x`. The Sturges estimator is quite good for small (<1000) datasets and is the default in the R language. This method gives good off the shelf behaviour. Parameters ---------- x : array_like Input data that is to be histogrammed, trimmed to range. May not be empty. Returns ------- h : An estimate of the optimal bin width for the given data. See Also -------- _hist_bin_fd, _hist_bin_sturges """ # There is no need to check for zero here. If ptp is, so is IQR and # vice versa. Either both are zero or neither one is. return min(_hist_bin_fd(x), _hist_bin_sturges(x)) ######################################################################
19,149
35.406844
97
py
GAStimator
GAStimator-master/gastimator/gastimator.py
#!/usr/bin/env python3 # coding: utf-8 import numpy as np import matplotlib.pyplot as plt from tqdm import tqdm from joblib import Parallel, delayed,cpu_count from joblib.externals.loky import get_reusable_executor def lnlike(data,model,error): # default log-likelihood function chi2=np.nansum((data - model)**2 / error**2) return -0.5*chi2 def unwrap_self(args, **kwarg): return gastimator.run_a_chain(*args, **kwarg) class gastimator: def __init__(self, model,*args, **kwargs): self.targetrate= 0.25 self.dec=0.95 self.inc=1.05 self.model=model self.silent=False self.rng=None self.args=args self.kwargs=kwargs self.min=None self.max=None self.guesses=None self.precision=None self.prior_func=None self.fixed=None self.fdata=None self.error=None self.labels=None self.change=None self.npars=None self.lastchain=None self.lastchainll=None self.lastchainaccept=None self.nprocesses=int(cpu_count())-1 self.lnlike_func=lnlike # def likelihood(self,values): # # priorval=1 # for ival,prior in enumerate(self.prior_func): # if callable(prior): # priorval*=prior(values[ival],allvalues=values,ival=ival) # # model=self.model(values,*self.args,**self.kwargs) # return self.lnlike_func(self.fdata,model,self.error) + np.log(priorval, where=(priorval!=0)) def likelihood(self,values): priorval=0 for ival,prior in enumerate(self.prior_func): if callable(prior): priorval+=prior(values[ival],allvalues=values,ival=ival) model=self.model(values,*self.args,**self.kwargs) return self.lnlike_func(self.fdata,model,self.error) + priorval def take_step(self,values,changethistime, knob): newvals=values.copy() acceptable=False cnt=0.0 while not acceptable: newvals[changethistime]=values[changethistime]+(self.rng.randn(1)*knob[changethistime]) if (newvals[changethistime] >= self.min[changethistime]) * (newvals[changethistime] <= self.max[changethistime]): acceptable=True if cnt > 5000: raise Exception('Cannot find a good step. Check input parameters') cnt+=1 return newvals def evaluate_step(self, oldvalues, oldll, changethistime, knob, holdknob=False): accept=True tryvals=self.take_step(oldvalues,changethistime, knob) newll=self.likelihood(tryvals) if np.isnan(newll): print("Your function returned a Nan for the following parameters",tryvals) print("Attempting to continue...") ll=-1e20 ratio= newll-oldll randacc=np.log(self.rng.uniform(0,1,1)) if randacc > ratio: accept=False newval = oldvalues newll = oldll if not holdknob: knob[changethistime]*=self.dec else: newval=tryvals if (not holdknob) and (knob[changethistime] < (self.max[changethistime]-self.min[changethistime])*5.): knob[changethistime]*=self.inc return newval, newll, accept, tryvals, knob def chain(self, values, niter, knob, plot=True, holdknob=False, fig=None, ax=None,progress=False,progid=0): if plot: fig, ax=self.make_plots(niter,progid=progid) #plt.pause(0.00001) gibbs_change=self.rng.randint(0,self.change.size,niter) outputvals=np.empty((self.npars,niter)) outputll=np.zeros(niter) accepted=np.zeros(niter) bestll=self.likelihood(values) outputvals[:,0]=values outputll[0]=bestll n_accept=0. if progress and (progid==0): pbar = tqdm(total=(niter)*self.nprocesses) for i in range(1,niter): newval, newll, accept, tryvals, knob =self.evaluate_step(outputvals[:,i-1],outputll[i-1],self.change[gibbs_change[i]],knob,holdknob=holdknob) if progress and (progid==0): pbar.update(self.nprocesses) if plot: self.update_plot(fig, ax,self.change[gibbs_change[i]],i,tryvals, accept) outputvals[:,i]=newval outputll[i]=newll accepted[i]=accept n_accept+=accept acceptrate=float(n_accept)/niter if (holdknob): if (self.burn): outputvals= outputvals[:,int(self.burn):-1] outputll= outputll[int(self.burn):-1] accepted=accepted[int(self.burn):-1] outputvals=outputvals[:,accepted.astype(bool)] outputll=outputll[accepted.astype(bool)] if plot: plt.close() if progress and (progid==0): pbar.close() return outputvals, outputll, accepted, acceptrate, knob def factors(self,n): facs=np.array(list(x for tup in ([i, n//i] for i in range(1, int(n**0.5)+1) if n % i == 0) for x in tup)) facs=np.sort(facs) return facs[int(facs.size/2 - 1):int(facs.size/2 +1)] def update_plot(self, fig, ax, index, i, new_data, accept): plot2update=(ax.flat[index]) if accept: plot2update.plot(i,new_data[index],'bo') else: plot2update.plot(i,new_data[index],'ro') plt.pause(0.000001) #fig.canvas.draw() #fig.canvas.start_event_loop(0.000001) def make_plots(self, niter,progid=0): nplots=self.npars plotfac=self.factors(nplots) fig, ax = plt.subplots(plotfac[1], plotfac[0], sharex='col') for i in range(0,nplots): (ax.flat[i]).set_ylabel(self.labels[i]+str(progid)) (ax.flat[i]).set_xlim([0,niter]) (ax.flat[i]).set_ylim([self.min[i],self.max[i]]) plt.show(block=False) return fig, ax def run_a_chain(self,startguess,niters,numatonce,knob,plot=True,final=False,progid=0): count=0 converged=False oldmean=self.guesses*0.0 newmean=0.0 startpoint=startguess.copy() self.rng=np.random.RandomState() if not final: while (count < niters) and (not converged): outputvals, outputll, accepted, acceptrate, knob = self.chain(startpoint, numatonce, knob, plot=plot, holdknob=False,progress=False) if acceptrate*numatonce > 1: w,=np.where(accepted) startpoint=outputvals[:,w[-1]] newmean=np.mean(outputvals[:,w],axis=1) else: newmean=oldmean if count == 0: self.lastchain=outputvals self.lastchainll=outputll self.lastchainaccept=accepted if (not self.silent): print(" Chain "+str(progid)+" has not converged - Accept rate: "+str(acceptrate)) else: self.lastchain=np.append(self.lastchain,outputvals,axis=1) self.lastchainll=np.append(self.lastchainll,outputll) self.lastchainaccept=np.append(self.lastchainaccept,accepted) test=(np.abs(newmean-oldmean) < self.precision) if test[self.change].sum() == test[self.change].size and (acceptrate >= self.targetrate): converged=1 if (not self.silent): print("Chain "+str(progid)+" converged: LL: "+str(np.max(outputll))+" - Accept rate:"+str(acceptrate)) else: if (not self.silent): print(" Chain "+str(progid)+" has not converged - Accept rate: "+str(acceptrate)) if test[self.change].sum() == test[self.change].size: if (not self.silent): print(" --> Target rate not reached") else: if (not self.silent): print(" --> Still varying: "+str(self.labels[self.change][~test[self.change]])) oldmean=newmean count += numatonce if not converged: print('WARNING: Chain '+str(progid)+' did not converge in '+str(niters)+' steps') else: #if (self.silent == False)&(self.nprocesses==1): # outputvals, outputll, accepted, acceptrate, knob = self.chain(startpoint, niters, knob, plot=False, holdknob=True,progress=True,progid=progid) #else: #self.rng=np.random.RandomState() #refresh the RNG outputvals, outputll, accepted, acceptrate, knob = self.chain(startpoint, niters, knob, plot=False, holdknob=True,progress=True,progid=progid) best_knob=knob return outputvals, outputll, best_knob def input_checks(self): self.guesses=np.array(self.guesses) self.min=np.array(self.min) self.max=np.array(self.max) self.precision=np.array(self.precision) self.labels=np.array(self.labels) if np.any(self.guesses == None): raise Exception('Please set initial guesses') self.npars=self.guesses.size if np.all(np.array(self.prior_func) == None): self.prior_func=(np.zeros(self.npars, dtype=bool)).tolist() else: try: if len(self.prior_func) != self.npars: raise Exception('Number of priors given does not match number of parameters') except: if self.npars != 1: raise Exception('Number of priors given does not match number of parameters') names=["minimum","maximum","precision","labels"] check=[self.min,self.max,self.precision,self.labels] for x, nam in zip(check,names): if np.any(x == None): raise Exception('Please set parameter '+str(nam)) else: if x.size != self.npars: raise Exception('Number of constraints in '+str(nam)+' does not match number of parameters') if np.any(self.fixed == None): self.fixed=np.zeros(self.npars, dtype=bool) print("You did not specify if any variables are fixed - I will continue assuming that none are") if np.any((self.max-self.min) < 0): raise Exception('Parameter(s) '+str(self.labels[(self.max-self.min) < 0])+' have incorrect minumum/maximum bounds') if np.any((self.guesses<self.min)): raise Exception('Parameter(s) '+str(self.labels[(self.guesses<self.min)])+' have an initial guess ('+str(self.guesses[(self.guesses<self.min)])+') lower than the minimum allowed ('+str(self.min[(self.guesses<self.min)])+').') if np.any((self.guesses>self.max)): raise Exception('Parameter(s) '+str(self.labels[(self.guesses>self.max)])+' have an initial guess higher ('+str(self.guesses[(self.guesses>self.max)])+') than the maximum allowed ('+str(self.max[(self.guesses>self.max)])+').') def run(self, fdata, error, niters, numatonce=None, burn=None, nchains=1, plot=True, output=None, seed=None): # check all required inputs set self.input_checks() if not self.silent: stri='' for name,pr in zip(self.labels,self.prior_func): if callable(pr): stri=stri+' '+name if stri!='': print("Non uniform priors used: "+stri) # set up variables needed self.fdata=fdata self.error=error self.rng=np.random.RandomState(seed) self.change, = np.where(self.fixed == False) verybestvalues=self.guesses verybestknob=None verybestll=-1e31 if ((0.8*(niters/self.nprocesses)) < 1000) & (self.nprocesses > 1): print("WARNING: The chain assigned to each processor will be very short (<1250 steps) - consider reducing 'nprocesses'.") if not numatonce: numatonce=50*self.npars if not burn: self.burn=0.2*(niters/self.nprocesses) else: self.burn=burn if (self.silent==False)&(self.nprocesses>1): verboselev=10 else: verboselev=0 self.progline=[] for chainno in range(0,nchains): self.progline.append('Doing chain '+str(chainno+1)) knob=(0.5*(self.max-self.min)) #breakpoint() if (self.nprocesses>1)&(nchains>1): ## do initial chains in parallel #breakpoint() #try: par= Parallel(n_jobs= self.nprocesses, verbose=verboselev) results=par(delayed(unwrap_self)(i) for i in zip([self]*nchains, [self.guesses]*nchains,[int(float(niters))]*nchains,[numatonce]*nchains,[knob]*nchains, [plot]*nchains, [False]*nchains,np.arange(nchains))) #except: # par= Parallel(n_jobs= self.nprocesses, verbose=verboselev, prefer="threads") # results=par(delayed(unwrap_self)(i) for i in zip([self]*nchains, [self.guesses]*nchains,[int(float(niters))]*nchains,[numatonce]*nchains,[knob]*nchains, [plot]*nchains, [False]*nchains,np.arange(nchains))) results=np.array(results,dtype=object) par._terminate_backend() get_reusable_executor().shutdown(wait=True) ## #debug line #results=unwrap_self((self, self.guesses,int(float(niters)),numatonce,knob, False, False,0)) bestchain=np.argmax(np.max(np.stack(results[:,1]),axis=1)) bestvalinbestchain=np.argmax(results[bestchain,1]) verybestvalues=np.stack(results[bestchain,0])[:,bestvalinbestchain] verybestknob=results[bestchain,2] verybestll=np.stack(results[bestchain,1])[bestvalinbestchain] else: for chainno in range(0,nchains): if not self.silent: print('Doing chain '+str(chainno+1)) #knob=(0.5*(self.max-self.min)) outputvals, outputll, best_knob = self.run_a_chain(self.guesses,niters,numatonce,knob,plot=plot) #breakpoint() if (np.max(outputll) > verybestll) or (chainno == 0): if not self.silent: print("Best chain so far!") w,=np.where(outputll == np.max(outputll)) verybestvalues=outputvals[:,w[0]].reshape(self.npars) verybestknob=best_knob verybestll=np.max(outputll) if not self.silent: print("Best fit:") for i in range(0,self.labels.size): print(" "+self.labels[i]+":",verybestvalues[i]) print("Starting final chain") results = [] #try: par= Parallel(n_jobs= self.nprocesses, verbose=verboselev) results=par(delayed(unwrap_self)(i) for i in zip([self]*self.nprocesses, [verybestvalues]*self.nprocesses,[int(float(niters)/float(self.nprocesses))]*self.nprocesses,[numatonce]*self.nprocesses,[verybestknob]*self.nprocesses, [False]*self.nprocesses, [True]*self.nprocesses,np.arange(self.nprocesses))) # except: # par= Parallel(n_jobs= self.nprocesses, verbose=verboselev,prefer="threads") # results=par(delayed(unwrap_self)(i) for i in zip([self]*self.nprocesses, [verybestvalues]*self.nprocesses,[int(float(niters)/float(self.nprocesses))]*self.nprocesses,[numatonce]*self.nprocesses,[verybestknob]*self.nprocesses, [False]*self.nprocesses, [True]*self.nprocesses,np.arange(self.nprocesses))) results=np.array(results,dtype=object) par._terminate_backend() get_reusable_executor().shutdown(wait=True) outputvalue= np.concatenate(results[:,0],axis=1) outputll= np.concatenate(results[:,1]) if outputll.size < 1: print('WARNING: No accepted models. Perhaps you need to increase the number of steps?') else: if not self.silent: w,=np.where(outputll == np.max(outputll)) perc = np.percentile(outputvalue, [15.86, 50, 84.14], axis=1) sig_bestfit_up = (perc[2][:] - perc[1][:]) sig_bestfit_down = (perc[1][:] - perc[0][:]) print("Final best fit values and 1sigma errors:") for i in range(0,self.labels.size): if self.fixed[i]: print(" "+self.labels[i]+":",perc[1][i],"(Fixed)") else: if np.abs(((sig_bestfit_up[i]/sig_bestfit_down[i])-1)) < 0.1: print(" "+self.labels[i]+":",perc[1][i],"±",np.mean([sig_bestfit_up[i],sig_bestfit_down[i]])) else: print(" "+self.labels[i]+":",perc[1][i],"+",sig_bestfit_up[i],"-",sig_bestfit_down[i]) return outputvalue, outputll
16,793
38.146853
312
py
python-pesq
python-pesq-master/setup.py
from setuptools import setup, Extension, find_packages from setuptools.command.build_ext import build_ext as _build_ext import os includes = ['pypesq'] try: import numpy as np includes += [os.path.join(np.get_include(), 'numpy')] except: pass extension = Extension("pesq_core", sources=["pypesq/pesq.c", "pypesq/dsp.c", "pypesq/pesqdsp.c", "pypesq/pesqio.c", "pypesq/pesqmain.c", "pypesq/pesqmod.c"], include_dirs=includes, language='c') class build_ext(_build_ext): def finalize_options(self): _build_ext.finalize_options(self) try: __builtins__.__NUMPY_SETUP__ = False except AttributeError: print("Cannot set '__builtins__.__NUMPY_SETUP__ = False' This is not needed if numpy is already installed.") import numpy self.include_dirs.append(numpy.get_include()) setup(name='pypesq', version='1.2.4', description="A package to compute pesq score.", url='https://github.com/vBaiCai/python-pesq', author_email='zhuroubaicai@gmail.com', keywords=['pesq', 'speech', 'speech quality'], license='MIT', packages=find_packages(), ext_modules=[extension], cmdclass={'build_ext': build_ext}, setup_requires=['numpy'], py_modules=['numpy'], zip_safe=False, install_requires=['numpy'], python_requires='!=3.0.*, !=3.1.*, !=3.2.*, <4', )
1,498
30.893617
120
py
python-pesq
python-pesq-master/pypesq/__init__.py
import warnings import numpy as np from pesq_core import _pesq from math import fabs EPSILON = 1e-6 def pesq(ref, deg, fs=16000, normalize=False): ''' params: ref: ref signal, deg: deg signal, fs: sample rate, ''' ref = np.array(ref, copy=True) deg = np.array(deg, copy=True) if normalize: ref = ref/np.max(np.abs(ref)) if np.abs(ref) > EPSILON else ref deg = deg/np.max(np.abs(deg)) if np.abs(deg) > EPSILON else deg max_sample = np.max(np.abs(np.array([ref, deg]))) if max_sample > 1: c = 1 / max_sample ref = ref * c deg = deg * c if ref.ndim != 1 or deg.ndim != 1: raise ValueError("signals must be 1-D array ") if fs not in [16000, 8000]: raise ValueError("sample rate must be 16000 or 8000") if fabs(ref.shape[0] - deg.shape[0]) > fs / 4: raise ValueError("ref and deg signals should be in same length.") if np.count_nonzero(ref==0) == ref.size: raise ValueError("ref is all zeros, processing error! ") if np.count_nonzero(deg==0) == deg.size: raise ValueError("deg is all zeros, pesq score is nan! ") if ref.dtype != np.int16: ref *= 32767 ref = ref.astype(np.int16) if deg.dtype != np.int16: deg *= 32767 deg = deg.astype(np.int16) try: score = _pesq(ref, deg, fs) except: warnings.warn('Processing Error! return NaN') score = np.NAN return score
1,501
25.350877
73
py
scaper
scaper-master/setup.py
from setuptools import setup import imp with open('README.md') as file: long_description = file.read() version = imp.load_source('scaper.version', 'scaper/version.py') setup( name='scaper', version=version.version, description='A library for soundscape synthesis and augmentation', author='Justin Salamon & Duncan MacConnell', author_email='justin.salamon@gmail.com', url='https://github.com/justinsalamon/scaper', download_url='http://github.com/justinsalamon/scaper/releases', packages=['scaper'], long_description=long_description, long_description_content_type='text/markdown', keywords='audio sound soundscape environmental dsp mixing', license='BSD-3-Clause', classifiers=[ "License :: OSI Approved :: BSD License", "Programming Language :: Python", "Development Status :: 3 - Alpha", "Intended Audience :: Developers", "Intended Audience :: Science/Research", "Topic :: Multimedia :: Sound/Audio :: Analysis", "Topic :: Multimedia :: Sound/Audio :: Sound Synthesis", "Programming Language :: Python :: 2", "Programming Language :: Python :: 2.7", "Programming Language :: Python :: 3", "Programming Language :: Python :: 3.4", "Programming Language :: Python :: 3.5", "Programming Language :: Python :: 3.6", ], install_requires=[ 'sox==1.4.0', 'jams>=0.3.2', 'numpy>=1.13.3', "soxbindings>=1.2.2;platform_system!='Windows'", 'pyloudnorm', 'soundfile', ], extras_require={ 'docs': [ 'sphinx', # autodoc was broken in 1.3.1 'sphinx_rtd_theme', 'sphinx_issues', ], 'tests': ['backports.tempfile', 'pytest', 'pytest-cov', 'tqdm'] } )
1,903
33.618182
71
py
scaper
scaper-master/tests/test_core.py
import scaper from scaper.scaper_exceptions import ScaperError from scaper.scaper_warnings import ScaperWarning from scaper.util import _close_temp_files import pytest from scaper.core import EventSpec import tempfile import backports.tempfile import os import numpy as np import soundfile import jams import numbers from collections import namedtuple from copy import deepcopy import shutil from contextlib import contextmanager import csv # FIXTURES # Paths to files for testing FG_PATH = 'tests/data/audio/foreground' BG_PATH = 'tests/data/audio/background' SHORT_BG_PATH = 'tests/data/audio/short_background' ALT_FG_PATH = 'tests/data/audio_alt_path/foreground' ALT_BG_PATH = 'tests/data/audio_alt_path/background' # fg and bg labels for testing FB_LABELS = ['car_horn', 'human_voice', 'siren'] BG_LABELS = ['park', 'restaurant', 'street'] _EXTS = ('wav', 'jams', 'txt') _TestFiles = namedtuple('TestFiles', _EXTS) SAMPLE_RATES = [44100, 22050] def _get_test_paths(name): return _TestFiles(*[ os.path.join('tests/data/regression/', name + '.' + ext) for ext in _EXTS ]) TEST_PATHS = { 22050: { 'REG': _get_test_paths('soundscape_20200501_22050'), 'REG_BGONLY': _get_test_paths('bgonly_soundscape_20200501_22050'), 'REG_REVERB': _get_test_paths('reverb_soundscape_20200501_22050'), }, 44100: { 'REG': _get_test_paths('soundscape_20200501_44100'), 'REG_BGONLY': _get_test_paths('bgonly_soundscape_20200501_44100'), 'REG_REVERB': _get_test_paths('reverb_soundscape_20200501_44100'), }, } def _compare_scaper_jams(jam, regjam, exclude_additional_scaper_sandbox_keys=[]): """ Check whether two scaper jams objects are equal up to floating point precision, ignoring jams_version and scaper_version. Parameters ---------- jam : JAMS In memory jams object regjam : JAMS Regression jams (loaded from disk) Raises ------ AssertionError If the comparison fails. """ # Note: can't compare directly, since: # 1. scaper/and jams library versions may change # 2. raw annotation sandbox stores specs as OrderedDict and tuples, whereas # loaded ann (regann) simplifies those to dicts and lists # 3. floats might be marginally different (need to use np.allclose()) # Must compare each part "manually" # 1. compare file metadata for k in set(jam.file_metadata.keys()) | set(regjam.file_metadata.keys()): if k != 'jams_version': assert jam.file_metadata[k] == regjam.file_metadata[k] # 2. compare jams sandboxes assert jam.sandbox == regjam.sandbox # 3. compare annotations assert len(jam.annotations) == len(regjam.annotations) == 1 ann = jam.annotations[0] regann = regjam.annotations[0] # 3.1 compare annotation metadata assert ann.annotation_metadata == regann.annotation_metadata # 3.2 compare sandboxes # Note: can't compare sandboxes directly, since in raw jam scaper sandbox # stores event specs in EventSpec object (named tuple), whereas in loaded # jam these will get converted to list of lists. # assert ann.sandbox == regann.sandbox assert len(ann.sandbox.keys()) == len(regann.sandbox.keys()) == 1 assert 'scaper' in ann.sandbox.keys() assert 'scaper' in regann.sandbox.keys() excluded_scaper_sandbox_keys = [ 'bg_spec', 'fg_spec', 'scaper_version', 'soundscape_audio_path', 'isolated_events_audio_path', ] excluded_scaper_sandbox_keys.extend(exclude_additional_scaper_sandbox_keys) # everything but the specs and version can be compared directly: for k in set(ann.sandbox.scaper.keys()) | set(regann.sandbox.scaper.keys()): if k not in excluded_scaper_sandbox_keys: assert ann.sandbox.scaper[k] == regann.sandbox.scaper[k], ( 'Unequal values for "{}"'.format(k)) # to compare specs need to covert raw specs to list of lists bg_spec_list = [[list(x) if isinstance(x, tuple) else x for x in e] for e in ann.sandbox.scaper['bg_spec']] fg_spec_list = [[list(x) if isinstance(x, tuple) else x for x in e] for e in ann.sandbox.scaper['fg_spec']] assert (fg_spec_list == regann.sandbox.scaper['fg_spec']) assert (bg_spec_list == regann.sandbox.scaper['bg_spec']) # 3.3. compare namespace, time and duration assert ann.namespace == regann.namespace assert ann.time == regann.time assert ann.duration == regann.duration # 3.4 compare data for obs, regobs in zip(ann.data, regann.data): # compare time, duration and confidence assert np.allclose(obs.time, regobs.time) assert np.allclose(obs.duration, regobs.duration) assert np.allclose(obs.confidence, regobs.confidence) # compare value dictionaries v, regv = obs.value, regobs.value for k in set(v.keys()) | set(regv.keys()): if isinstance(v[k], numbers.Number): assert np.allclose(v[k], regv[k]) else: assert v[k] == regv[k] def _compare_txt_annotation(orig_txt_file, gen_txt_file): # if string - read in both files. # if list, assume data already loaded if type(orig_txt_file) is str: txt_data = _load_txt_annotation(orig_txt_file) else: txt_data = orig_txt_file if type(gen_txt_file) is str: gen_txt_data = _load_txt_annotation(gen_txt_file) else: gen_txt_data = gen_txt_file # compare start and end times for txt_row, gen_txt_row in zip(txt_data, gen_txt_data): assert np.allclose(txt_row[:2], gen_txt_row[:2]) # compare times assert txt_row[2] == gen_txt_row[2] # compare labels def _load_txt_annotation(txt_file, delimiter='\t'): """ Load a tab-delimited simplified annotation as a list of lists. Assumes first two elements are floats (start/end times) and the remaining item is a string (the label). Parameters ---------- txt_file : str Path to simplified annotation txt file. Returns ------- data : list List of annotation data, each row is one sound event represented as: [start_time, end_time, label]. """ data = [] with open(txt_file, 'r') as f: reader = csv.reader(f, delimiter=delimiter) for row in reader: row = [float(x) if n <= 1 else x for n, x in enumerate(row)] data.append(row) return data def test_generate_from_jams(atol=1e-5, rtol=1e-8): # Test for invalid jams: no annotations tmpfiles = [] with _close_temp_files(tmpfiles): jam = jams.JAMS() jam.file_metadata.duration = 10 jam_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True) gen_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True) jam.save(jam_file.name) pytest.raises(ScaperError, scaper.generate_from_jams, jam_file.name, gen_file.name) # # Make sure we can load an old JAM file that doesn't have fix_cilpping or peak_normalization old_jam_file = 'tests/data/regression/soundscape_20200501_44100_no_clipping_normalization_fields.jams' tmpfiles = [] with _close_temp_files(tmpfiles): gen_audio_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) gen_jam_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True) gen_txt_file = tempfile.NamedTemporaryFile(suffix='.txt', delete=True) tmpfiles.extend([gen_audio_file, gen_jam_file, gen_txt_file]) (fj_soundscape_audio, fj_soundscape_jam, fj_annotation_list, fj_event_audio_list) = \ scaper.generate_from_jams(old_jam_file, audio_outfile=gen_audio_file.name, jams_outfile=gen_jam_file.name, txt_path=gen_txt_file.name) # validate return API orig_wav, sr = soundfile.read(TEST_PATHS[44100]['REG'].wav, always_2d=True) assert np.allclose(orig_wav, fj_soundscape_audio) regjam = jams.load(TEST_PATHS[44100]['REG'].jams) sandbox_exclude = ['fix_clipping', 'peak_normalization', 'quick_pitch_time'] _compare_scaper_jams( regjam, fj_soundscape_jam, exclude_additional_scaper_sandbox_keys=sandbox_exclude) # _compare_txt_annotation(annotation_list, fj_annotation_list) # TODO # TODO: # for event, fj_event in zip(event_audio_list, fj_event_audio_list): # assert np.allclose(event, fj_event, atol=1e-8, rtol=rtol) # validate soundscape audio written to disk gen_wav, sr = soundfile.read(gen_audio_file.name, always_2d=True) assert np.allclose(gen_wav, orig_wav, atol=atol, rtol=rtol) # validate jams sandbox_exclude = ['fix_clipping', 'peak_normalization', 'quick_pitch_time'] gen_jam = jams.load(gen_jam_file.name) _compare_scaper_jams( regjam, gen_jam, exclude_additional_scaper_sandbox_keys=sandbox_exclude) # validate annotation txt _compare_txt_annotation(TEST_PATHS[44100]['REG'].txt, gen_txt_file.name) # Test for valid jams file tmpfiles = [] with _close_temp_files(tmpfiles): # Create all necessary temp files orig_wav_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) orig_jam_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True) orig_txt_file = tempfile.NamedTemporaryFile(suffix='.txt', delete=True) gen_wav_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) gen_jam_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True) gen_txt_file = tempfile.NamedTemporaryFile(suffix='.txt', delete=True) tmpfiles.append(orig_wav_file) tmpfiles.append(orig_jam_file) tmpfiles.append(orig_txt_file) tmpfiles.append(gen_wav_file) tmpfiles.append(gen_jam_file) tmpfiles.append(gen_txt_file) # --- Define scaper --- * sc = scaper.Scaper(10, FG_PATH, BG_PATH) sc.protected_labels = [] sc.ref_db = -50 sc.add_background(label=('choose', []), source_file=('choose', []), source_time=('const', 0)) # Add 5 events for _ in range(5): sc.add_event(label=('choose', []), source_file=('choose', []), source_time=('const', 0), event_time=('uniform', 0, 9), event_duration=('choose', [1, 2, 3]), snr=('uniform', 10, 20), pitch_shift=('uniform', -1, 1), time_stretch=('uniform', 0.8, 1.2)) # --- Define CLIPPING scaper --- * sc_clipping = scaper.Scaper(10, FG_PATH, BG_PATH) sc_clipping.protected_labels = [] sc_clipping.ref_db = 0 sc_clipping.add_background(label=('choose', []), source_file=('choose', []), source_time=('const', 0)) # Add 5 events for _ in range(5): sc_clipping.add_event(label=('choose', []), source_file=('choose', []), source_time=('const', 0), event_time=('uniform', 0, 9), event_duration=('choose', [1, 2, 3]), snr=('uniform', 20, 30), pitch_shift=('uniform', -1, 1), time_stretch=('uniform', 0.8, 1.2)) def _validate_soundscape_and_event_audio(orig_wav_file, gen_wav_file, gen_events_path, orig_events_path): # validate audio orig_wav, sr = soundfile.read(orig_wav_file.name) gen_wav, sr = soundfile.read(gen_wav_file.name) assert np.allclose(gen_wav, orig_wav, atol=atol, rtol=rtol) # validate that the sum of event audio sums to trimmed soundscape gen_event_files = [ os.path.join(gen_events_path, x) for x in sorted(os.listdir(gen_events_path)) ] gen_audio = [soundfile.read(x)[0] for x in gen_event_files] # Trim does not currently support trimming isolated events, but if/when # we add that functionality, this test should be updated to test that # as well, using the files in orig_events_path (currently unused). assert np.allclose(gen_wav, sum(gen_audio), atol=1e-8, rtol=rtol) # generate, then generate from the jams and compare audio files # repeat 5 times for _ in range(5): (soundscape_audio, soundscape_jam, annotation_list, event_audio_list) = \ sc.generate(audio_path=orig_wav_file.name, jams_path=orig_jam_file.name, txt_path=orig_txt_file.name, disable_instantiation_warnings=True) (fj_soundscape_audio, fj_soundscape_jam, fj_annotation_list, fj_event_audio_list) = \ scaper.generate_from_jams(orig_jam_file.name, audio_outfile=gen_wav_file.name, jams_outfile=gen_jam_file.name, txt_path=gen_txt_file.name) # validate return API assert np.allclose(soundscape_audio, fj_soundscape_audio) _compare_scaper_jams(soundscape_jam, fj_soundscape_jam) _compare_txt_annotation(annotation_list, fj_annotation_list) for event, fj_event in zip(event_audio_list, fj_event_audio_list): assert np.allclose(event, fj_event, atol=1e-8, rtol=rtol) # validate soundscape audio orig_wav, sr = soundfile.read(orig_wav_file.name) gen_wav, sr = soundfile.read(gen_wav_file.name) assert np.allclose(gen_wav, orig_wav, atol=atol, rtol=rtol) # validate jams orig_jam = jams.load(orig_jam_file.name) gen_jam = jams.load(gen_jam_file.name) _compare_scaper_jams(orig_jam, gen_jam) # validate annotation txt _compare_txt_annotation(orig_txt_file.name, gen_txt_file.name) # Test when we generate ONLY a JAMS file, and then generate audio from the JAMS # Case 1: without clipping for _ in range(5): (soundscape_audio, soundscape_jam, annotation_list, event_audio_list) = \ sc.generate(audio_path=orig_wav_file.name, jams_path=orig_jam_file.name, txt_path=orig_txt_file.name, no_audio=True, disable_instantiation_warnings=True) assert soundscape_audio is None assert event_audio_list is None assert soundscape_jam is not None assert annotation_list is not None ann = soundscape_jam.annotations.search(namespace='scaper')[0] assert ann.sandbox.scaper.audio_path == orig_wav_file.name assert ann.sandbox.scaper.jams_path == orig_jam_file.name assert ann.sandbox.scaper.fix_clipping is False assert ann.sandbox.scaper.peak_normalization is False assert ann.sandbox.scaper.quick_pitch_time is False assert ann.sandbox.scaper.save_isolated_events is False assert ann.sandbox.scaper.isolated_events_path is None assert ann.sandbox.scaper.disable_sox_warnings is True assert ann.sandbox.scaper.no_audio is True assert ann.sandbox.scaper.txt_path == orig_txt_file.name assert ann.sandbox.scaper.txt_sep is '\t' assert ann.sandbox.scaper.disable_instantiation_warnings is True assert ann.sandbox.scaper.peak_normalization_scale_factor == 1.0 assert ann.sandbox.scaper.ref_db_change == 0 assert ann.sandbox.scaper.ref_db_generated == \ ann.sandbox.scaper.ref_db (fj_soundscape_audio, fj_soundscape_jam, fj_annotation_list, fj_event_audio_list) = \ scaper.generate_from_jams(orig_jam_file.name, audio_outfile=gen_wav_file.name, jams_outfile=gen_jam_file.name, txt_path=gen_txt_file.name) # validate return API _compare_scaper_jams(soundscape_jam, fj_soundscape_jam) _compare_txt_annotation(annotation_list, fj_annotation_list) # Test when we generate ONLY a JAMS file, and then generate audio from the JAMS # Case 2: WITH CLIPPING for _ in range(5): (soundscape_audio, soundscape_jam, annotation_list, event_audio_list) = \ sc_clipping.generate(audio_path=orig_wav_file.name, jams_path=orig_jam_file.name, txt_path=orig_txt_file.name, no_audio=True, fix_clipping=True, disable_instantiation_warnings=True) assert soundscape_audio is None assert event_audio_list is None assert soundscape_jam is not None assert annotation_list is not None ann = soundscape_jam.annotations.search(namespace='scaper')[0] assert ann.sandbox.scaper.audio_path == orig_wav_file.name assert ann.sandbox.scaper.jams_path == orig_jam_file.name assert ann.sandbox.scaper.fix_clipping is True assert ann.sandbox.scaper.peak_normalization is False assert ann.sandbox.scaper.quick_pitch_time is False assert ann.sandbox.scaper.save_isolated_events is False assert ann.sandbox.scaper.isolated_events_path is None assert ann.sandbox.scaper.disable_sox_warnings is True assert ann.sandbox.scaper.no_audio is True assert ann.sandbox.scaper.txt_path == orig_txt_file.name assert ann.sandbox.scaper.txt_sep is '\t' assert ann.sandbox.scaper.disable_instantiation_warnings is True assert ann.sandbox.scaper.peak_normalization_scale_factor == 1.0 assert ann.sandbox.scaper.ref_db_change == 0 assert ann.sandbox.scaper.ref_db_generated == \ ann.sandbox.scaper.ref_db (fj_soundscape_audio, fj_soundscape_jam, fj_annotation_list, fj_event_audio_list) = \ scaper.generate_from_jams(orig_jam_file.name, audio_outfile=gen_wav_file.name, jams_outfile=gen_jam_file.name, txt_path=gen_txt_file.name) assert fj_soundscape_audio is not None assert fj_event_audio_list is not None assert fj_soundscape_jam is not None assert fj_annotation_list is not None ann = fj_soundscape_jam.annotations.search(namespace='scaper')[0] # assert ann.sandbox.scaper.audio_path == gen_wav_file.name # assert ann.sandbox.scaper.jams_path == gen_jam_file.name assert ann.sandbox.scaper.fix_clipping is True assert ann.sandbox.scaper.peak_normalization is False assert ann.sandbox.scaper.quick_pitch_time is False assert ann.sandbox.scaper.save_isolated_events is False assert ann.sandbox.scaper.isolated_events_path is None assert ann.sandbox.scaper.disable_sox_warnings is True assert ann.sandbox.scaper.no_audio is True # TODO # assert ann.sandbox.scaper.txt_path == gen_txt_file.name assert ann.sandbox.scaper.txt_sep is '\t' assert ann.sandbox.scaper.disable_instantiation_warnings is True assert ann.sandbox.scaper.peak_normalization_scale_factor != 1.0 assert ann.sandbox.scaper.ref_db_change != 0 assert ann.sandbox.scaper.ref_db_generated != \ ann.sandbox.scaper.ref_db # Test when we generate ONLY a JAMS file, and then generate audio from the JAMS # Case 3: WITH quick_pitch_time=True, no clipping for _ in range(5): (soundscape_audio, soundscape_jam, annotation_list, event_audio_list) = \ sc.generate(audio_path=orig_wav_file.name, jams_path=orig_jam_file.name, txt_path=orig_txt_file.name, no_audio=True, fix_clipping=True, quick_pitch_time=True, disable_instantiation_warnings=True) assert soundscape_audio is None assert event_audio_list is None assert soundscape_jam is not None assert annotation_list is not None ann = soundscape_jam.annotations.search(namespace='scaper')[0] assert ann.sandbox.scaper.audio_path == orig_wav_file.name assert ann.sandbox.scaper.jams_path == orig_jam_file.name assert ann.sandbox.scaper.fix_clipping is True assert ann.sandbox.scaper.peak_normalization is False assert ann.sandbox.scaper.quick_pitch_time is True assert ann.sandbox.scaper.save_isolated_events is False assert ann.sandbox.scaper.isolated_events_path is None assert ann.sandbox.scaper.disable_sox_warnings is True assert ann.sandbox.scaper.no_audio is True assert ann.sandbox.scaper.txt_path == orig_txt_file.name assert ann.sandbox.scaper.txt_sep is '\t' assert ann.sandbox.scaper.disable_instantiation_warnings is True assert ann.sandbox.scaper.peak_normalization_scale_factor == 1.0 assert ann.sandbox.scaper.ref_db_change == 0 assert ann.sandbox.scaper.ref_db_generated == \ ann.sandbox.scaper.ref_db (fj_soundscape_audio, fj_soundscape_jam, fj_annotation_list, fj_event_audio_list) = \ scaper.generate_from_jams(orig_jam_file.name, audio_outfile=gen_wav_file.name, jams_outfile=gen_jam_file.name, txt_path=gen_txt_file.name) assert fj_soundscape_audio is not None assert fj_event_audio_list is not None assert fj_soundscape_jam is not None assert fj_annotation_list is not None ann = fj_soundscape_jam.annotations.search(namespace='scaper')[0] # assert ann.sandbox.scaper.audio_path == gen_wav_file.name # assert ann.sandbox.scaper.jams_path == gen_jam_file.name assert ann.sandbox.scaper.fix_clipping is True assert ann.sandbox.scaper.peak_normalization is False assert ann.sandbox.scaper.quick_pitch_time is True assert ann.sandbox.scaper.save_isolated_events is False assert ann.sandbox.scaper.isolated_events_path is None assert ann.sandbox.scaper.disable_sox_warnings is True assert ann.sandbox.scaper.no_audio is True # TODO # assert ann.sandbox.scaper.txt_path == gen_txt_file.name assert ann.sandbox.scaper.txt_sep is '\t' assert ann.sandbox.scaper.disable_instantiation_warnings is True assert ann.sandbox.scaper.peak_normalization_scale_factor == 1.0 assert ann.sandbox.scaper.ref_db_change == 0 assert ann.sandbox.scaper.ref_db_generated == \ ann.sandbox.scaper.ref_db # validate return API # _compare_scaper_jams(soundscape_jam, fj_soundscape_jam) # _compare_txt_annotation(annotation_list, fj_annotation_list) # Now add in trimming! for _ in range(5): with backports.tempfile.TemporaryDirectory() as isolated_events_path: orig_events_path = os.path.join(isolated_events_path, 'original') gen_events_path = os.path.join(isolated_events_path, 'generated') os.makedirs(orig_events_path) os.makedirs(gen_events_path) sc.generate(audio_path=orig_wav_file.name, jams_path=orig_jam_file.name, txt_path=orig_txt_file.name, disable_instantiation_warnings=True, save_isolated_events=True, isolated_events_path=orig_events_path) scaper.trim(orig_wav_file.name, orig_jam_file.name, orig_wav_file.name, orig_jam_file.name, np.random.uniform(0, 5), np.random.uniform(5, 10)) scaper.generate_from_jams(orig_jam_file.name, audio_outfile=gen_wav_file.name, txt_path=gen_txt_file.name, save_isolated_events=True, isolated_events_path=gen_events_path) _validate_soundscape_and_event_audio(orig_wav_file, gen_wav_file, gen_events_path, orig_events_path) # TODO # _compare_txt_annotation(orig_txt_file.name, gen_txt_file.name) # Double trimming for _ in range(2): with backports.tempfile.TemporaryDirectory() as isolated_events_path: orig_events_path = os.path.join(isolated_events_path, 'original') gen_events_path = os.path.join(isolated_events_path, 'generated') os.makedirs(orig_events_path) os.makedirs(gen_events_path) sc.generate(audio_path=orig_wav_file.name, jams_path=orig_jam_file.name, txt_path=orig_txt_file.name, disable_instantiation_warnings=True, save_isolated_events=True, isolated_events_path=orig_events_path) scaper.trim(orig_wav_file.name, orig_jam_file.name, orig_wav_file.name, orig_jam_file.name, np.random.uniform(0, 2), np.random.uniform(8, 10)) scaper.trim(orig_wav_file.name, orig_jam_file.name, orig_wav_file.name, orig_jam_file.name, np.random.uniform(0, 2), np.random.uniform(4, 6)) scaper.generate_from_jams(orig_jam_file.name, gen_wav_file.name, save_isolated_events=True, isolated_events_path=gen_events_path) _validate_soundscape_and_event_audio(orig_wav_file, gen_wav_file, gen_events_path, orig_events_path) # TODO # _compare_txt_annotation(orig_txt_file.name, gen_txt_file.name) # Triple trimming for _ in range(2): with backports.tempfile.TemporaryDirectory() as isolated_events_path: orig_events_path = os.path.join(isolated_events_path, 'original') gen_events_path = os.path.join(isolated_events_path, 'generated') os.makedirs(orig_events_path) os.makedirs(gen_events_path) sc.generate(audio_path=orig_wav_file.name, jams_path=orig_jam_file.name, txt_path=orig_txt_file.name, disable_instantiation_warnings=True, save_isolated_events=True, isolated_events_path=orig_events_path) scaper.trim(orig_wav_file.name, orig_jam_file.name, orig_wav_file.name, orig_jam_file.name, np.random.uniform(0, 2), np.random.uniform(8, 10)) scaper.trim(orig_wav_file.name, orig_jam_file.name, orig_wav_file.name, orig_jam_file.name, np.random.uniform(0, 1), np.random.uniform(5, 6)) scaper.trim(orig_wav_file.name, orig_jam_file.name, orig_wav_file.name, orig_jam_file.name, np.random.uniform(0, 1), np.random.uniform(3, 4)) scaper.generate_from_jams(orig_jam_file.name, gen_wav_file.name, save_isolated_events=True, isolated_events_path=gen_events_path) _validate_soundscape_and_event_audio(orig_wav_file, gen_wav_file, gen_events_path, orig_events_path) # TODO # _compare_txt_annotation(orig_txt_file.name, gen_txt_file.name) # Test with new FG and BG paths for _ in range(5): (soundscape_audio, soundscape_jam, annotation_list, event_audio_list) = \ sc.generate(audio_path=orig_wav_file.name, jams_path=orig_jam_file.name, txt_path=orig_txt_file.name, disable_instantiation_warnings=True) (fj_soundscape_audio, fj_soundscape_jam, fj_annotation_list, fj_event_audio_list) = \ scaper.generate_from_jams(orig_jam_file.name, audio_outfile=gen_wav_file.name, txt_path=gen_txt_file.name, fg_path=ALT_FG_PATH, bg_path=ALT_BG_PATH) # validate return API assert np.allclose(soundscape_audio, fj_soundscape_audio) # TODO: can't compare jams due to change in FG/BG # Need to be able to ignore specific keys in the event value dict # exclude_sandbox = ['fg_path', 'bg_path'] # _compare_scaper_jams( # soundscape_jam, fj_soundscape_jam, # exclude_additional_scaper_sandbox_keys=exclude_sandbox) _compare_txt_annotation(annotation_list, fj_annotation_list) for event, fj_event in zip(event_audio_list, fj_event_audio_list): assert np.allclose(event, fj_event, atol=1e-8, rtol=rtol) # validate soundscape audio orig_wav, sr = soundfile.read(orig_wav_file.name) gen_wav, sr = soundfile.read(gen_wav_file.name) assert np.allclose(gen_wav, orig_wav, atol=atol, rtol=rtol) # TODO: can't compare jams due to change in FG/BG # In the future update jam comparison such that any item can be # excluded from the comparison on demand, which would allow for this # test here. # orig_jam = jams.load(orig_jam_file.name) # gen_jam = jams.load(gen_jam_file.name) # _compare_scaper_jams(orig_jam, gen_jam) # validate annotation txt _compare_txt_annotation(orig_txt_file.name, gen_txt_file.name) # Ensure jam file saved correctly scaper.generate_from_jams(orig_jam_file.name, audio_outfile=gen_wav_file.name, jams_outfile=gen_jam_file.name) orig_jam = jams.load(orig_jam_file.name) gen_jam = jams.load(gen_jam_file.name) _compare_scaper_jams(orig_jam, gen_jam) def test_trim(atol=1e-5, rtol=1e-8): # Things we want to test: # 1. Jam trimmed correctly (mainly handled by jams.slice) # 2. value dict updated correctly (event_time, event_duration, source_time) # 3. scaper sandbox updated correctly (n_events, poly, gini, duration) # 4. audio trimmed correctly tmpfiles = [] with _close_temp_files(tmpfiles): # Create all necessary temp files orig_wav_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) orig_jam_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True) trim_wav_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) trim_jam_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True) trimstrict_wav_file = tempfile.NamedTemporaryFile( suffix='.wav', delete=True) trimstrict_jam_file = tempfile.NamedTemporaryFile( suffix='.jams', delete=True) tmpfiles.append(orig_wav_file) tmpfiles.append(orig_jam_file) tmpfiles.append(trim_wav_file) tmpfiles.append(trim_jam_file) tmpfiles.append(trimstrict_wav_file) tmpfiles.append(trimstrict_jam_file) # --- Create soundscape and save to tempfiles --- # sc = scaper.Scaper(10, FG_PATH, BG_PATH) sc.protected_labels = [] sc.ref_db = -50 sc.add_background(label=('const', 'park'), source_file=('choose', []), source_time=('const', 0)) # Add 5 events start_times = [0.5, 2.5, 4.5, 6.5, 8.5] for event_time in start_times: sc.add_event(label=('const', 'siren'), source_file=('choose', []), source_time=('const', 5), event_time=('const', event_time), event_duration=('const', 1), snr=('const', 10), pitch_shift=None, time_stretch=None) sc.generate(orig_wav_file.name, orig_jam_file.name, disable_instantiation_warnings=True) # --- Trim soundscape using scaper.trim with strict=False --- # scaper.trim(orig_wav_file.name, orig_jam_file.name, trim_wav_file.name, trim_jam_file.name, 3, 7, no_audio=False) # --- Validate output --- # # validate JAMS trimjam = jams.load(trim_jam_file.name) trimann = trimjam.annotations.search(namespace='scaper')[0] # Time and duration of annotation observation must be changed, but # values in the value dict must remained unchanged! for event in trimann.data: if event.value['role'] == 'background': assert (event.time == 0 and event.duration == 4 and event.value['event_time'] == 0 and event.value['event_duration'] == 10 and event.value['source_time'] == 0) else: if event.time == 0: assert (event.duration == 0.5 and event.value['event_time'] == 2.5 and event.value['event_duration'] == 1 and event.value['source_time'] == 5) elif event.time == 1.5: assert (event.duration == 1 and event.value['event_time'] == 4.5 and event.value['event_duration'] == 1 and event.value['source_time'] == 5) elif event.time == 3.5: assert (event.duration == 0.5 and event.value['event_time'] == 6.5 and event.value['event_duration'] == 1 and event.value['source_time'] == 5) else: assert False # validate audio orig_wav, sr = soundfile.read(orig_wav_file.name) trim_wav, sr = soundfile.read(trim_wav_file.name) assert np.allclose(trim_wav, orig_wav[3*sr:7*sr], atol=atol, rtol=rtol) def test_get_value_from_dist(): rng = scaper.util._check_random_state(0) # const x = scaper.core._get_value_from_dist(('const', 1), rng) assert x == 1 # choose for _ in range(10): x = scaper.core._get_value_from_dist(('choose', [1, 2, 3]), rng) assert x in [1, 2, 3] # uniform for _ in range(10): x = scaper.core._get_value_from_dist(('choose', [1, 2, 3]), rng) assert x in [1, 2, 3] # normal for _ in range(10): x = scaper.core._get_value_from_dist(('normal', 5, 1), rng) assert scaper.util.is_real_number(x) # truncnorm for _ in range(10): x = scaper.core._get_value_from_dist(('truncnorm', 5, 10, 0, 10), rng) assert scaper.util.is_real_number(x) assert 0 <= x <= 10 # COPY TESTS FROM test_validate_distribution (to ensure validation applied) def __test_bad_tuple_list(tuple_list): rng = scaper.util._check_random_state(0) for t in tuple_list: if isinstance(t, tuple): print(t, len(t)) else: print(t) pytest.raises(ScaperError, scaper.core._get_value_from_dist, t, random_state=rng) # not tuple = error nontuples = [[], 5, 'yes'] __test_bad_tuple_list(nontuples) # tuple must be at least length 2 shortuples = [tuple(), tuple(['const'])] __test_bad_tuple_list(shortuples) # unsupported distribution tuple name badnames = [('invalid', 1), ('constant', 1, 2, 3)] __test_bad_tuple_list(badnames) # supported dist tuples, but bad arugments badargs = [('const', 1, 2), ('choose', 1), ('choose', [], 1), ('uniform', 1), ('uniform', 1, 2, 3), ('uniform', 2, 1), ('uniform', 'one', 2), ('uniform', 1, 'two'), ('uniform', 0, 1j), ('uniform', 1j, 2), ('normal', 1), ('normal', 1, 2, 3), ('normal', 1, -1), ('normal', 0, 1j), ('normal', 1j, 1), ('normal', 'one', 2), ('normal', 1, 'two'), ('truncnorm', 1), ('truncnorm', 1, 2, 3), ('truncnorm', 1, -1, 0, 1), ('truncnorm', 0, 1j, 0, 1), ('truncnorm', 1j, 1, 0, 1), ('truncnorm', 'one', 2, 0, 1), ('truncnorm', 1, 'two', 0, 1), ('truncnorm', 1, 2, 'three', 5), ('truncnorm', 1, 2, 3, 'four'), ('truncnorm', 0, 2, 2, 0)] __test_bad_tuple_list(badargs) def test_ensure_satisfiable_source_time_tuple(): # Documenting the expected behavior of _ensure_satisfiable_source_time_tuple source_duration = 10 event_duration = 5 _test_dist = ('uniform', 4, 10) _adjusted, warn = scaper.core._ensure_satisfiable_source_time_tuple( _test_dist, source_duration, event_duration) assert (warn) assert np.allclose(_adjusted[1:], (4, 5)) _test_dist = ('truncnorm', 8, 1, 4, 10) _adjusted, warn = scaper.core._ensure_satisfiable_source_time_tuple( _test_dist, source_duration, event_duration) assert (warn) assert np.allclose(_adjusted[1:], (5, 1, 4, 5)) _test_dist = ('const', 6) _adjusted, warn = scaper.core._ensure_satisfiable_source_time_tuple( _test_dist, source_duration, event_duration) assert (warn) assert np.allclose(_adjusted[1:], (5)) _test_dist = ('uniform', 1, 10) _adjusted, warn = scaper.core._ensure_satisfiable_source_time_tuple( _test_dist, source_duration, event_duration) assert (warn) assert np.allclose(_adjusted[1:], (1, 5)) _test_dist = ('truncnorm', 4, 1, 1, 10) _adjusted, warn = scaper.core._ensure_satisfiable_source_time_tuple( _test_dist, source_duration, event_duration) assert (warn) assert np.allclose(_adjusted[1:], (4, 1, 1, 5)) _test_dist = ('uniform', 6, 10) _adjusted, warn = scaper.core._ensure_satisfiable_source_time_tuple( _test_dist, source_duration, event_duration) assert (warn) assert np.allclose(_adjusted[1], (5)) _test_dist = ('truncnorm', 8, 1, 6, 10) _adjusted, warn = scaper.core._ensure_satisfiable_source_time_tuple( _test_dist, source_duration, event_duration) assert (warn) assert np.allclose(_adjusted[1], (5)) _test_dist = ('choose', [0, 1, 2, 10, 12, 15, 20]) _adjusted, warn = scaper.core._ensure_satisfiable_source_time_tuple( _test_dist, source_duration, event_duration) assert (warn) assert np.allclose(_adjusted[1], [0, 1, 2, 5]) _test_dist = ('choose_weighted', [0, 1, 2, 10, 12, 15, 20], [0.1, 0.2, 0.05, 0.05, 0.3, 0.1, 0.2]) _adjusted, warn = scaper.core._ensure_satisfiable_source_time_tuple( _test_dist, source_duration, event_duration) assert (warn) assert np.allclose(_adjusted[1], [0, 1, 2, 5, 5, 5, 5]) assert np.allclose(_adjusted[2], [0.1, 0.2, 0.05, 0.05, 0.3, 0.1, 0.2]) def test_validate_distribution(): def __test_bad_tuple_list(tuple_list): for t in tuple_list: if isinstance(t, tuple): print(t, len(t)) else: print(t) pytest.raises(ScaperError, scaper.core._validate_distribution, t) # not tuple = error nontuples = [[], 5, 'yes'] __test_bad_tuple_list(nontuples) # tuple must be at least length 2 shortuples = [tuple(), tuple(['const'])] __test_bad_tuple_list(shortuples) # unsupported distribution tuple name badnames = [('invalid', 1), ('constant', 1, 2, 3)] __test_bad_tuple_list(badnames) # supported dist tuples, but bad arugments badargs = [('const', 1, 2), ('choose', 1), ('choose', [], 1), ('choose_weighted'), ('choose_weighted', []), ('choose_weighted', [1, 2, 3]), ('choose_weighted', [1, 2, 3], 5), ('choose_weighted', 5, [0.2, 0.3, 0.5]), ('choose_weighted', [1, 2, 3], []), ('choose_weighted', [1, 2, 3], [0.4, 0.6]), ('choose_weighted', [1, 2, 3], [0.4, 0.6, 0.1]), ('choose_weighted', [1, 2, 3], [0.4, 0.3, 0.2]), ('choose_weighted', [1, 2, 3], [0.7, -0.5, 0.8]), ('choose_weighted', [1, 2, 3], [1.2, -0.3, 0.1]), ('choose_weighted', [1, 2, 3], [1.2, 0.3, 0.1]), ('uniform', 1), ('uniform', 1, 2, 3), ('uniform', 2, 1), ('uniform', 'one', 2), ('uniform', 1, 'two'), ('uniform', 0, 1j), ('uniform', 1j, 2), ('normal', 1), ('normal', 1, 2, 3), ('normal', 1, -1), ('normal', 0, 1j), ('normal', 1j, 1), ('normal', 'one', 2), ('normal', 1, 'two'), ('truncnorm', 1), ('truncnorm', 1, 2, 3), ('truncnorm', 1, -1, 0, 1), ('truncnorm', 0, 1j, 0, 1), ('truncnorm', 1j, 1, 0, 1), ('truncnorm', 'one', 2, 0, 1), ('truncnorm', 1, 'two', 0, 1), ('truncnorm', 1, 2, 'three', 5), ('truncnorm', 1, 2, 3, 'four'), ('truncnorm', 0, 2, 2, 0)] __test_bad_tuple_list(badargs) def test_validate_label(): # label must be in list of allowed labels allowed_labels = ['yes'] pytest.raises(ScaperError, scaper.core._validate_label, ('const', 'no'), allowed_labels) # Choose list must be subset of allowed labels allowed_labels = ['yes', 'hello'] pytest.raises(ScaperError, scaper.core._validate_label, ('choose', ['no']), allowed_labels) # Label tuple must start with either 'const' or 'choose' bad_label_dists = [('uniform', 0, 1), ('normal', 0, 1), ('truncnorm', 0, 1, 0, 1)] for bld in bad_label_dists: pytest.raises(ScaperError, scaper.core._validate_label, bld, allowed_labels) def test_validate_source_file(): # file must exist # create temp folder so we have path to file we know doesn't exist with backports.tempfile.TemporaryDirectory() as tmpdir: nonfile = os.path.join(tmpdir, 'notafile') pytest.raises(ScaperError, scaper.core._validate_source_file, ('const', nonfile), ('const', 'siren')) # label must be const and match file foldername sourcefile = 'tests/data/audio/foreground/siren/69-Siren-1.wav' pytest.raises(ScaperError, scaper.core._validate_source_file, ('const', sourcefile), ('choose', [])) pytest.raises(ScaperError, scaper.core._validate_source_file, ('const', sourcefile), ('const', 'car_horn')) # if choose, all files in list of files must exist sourcefile = 'tests/data/audio/foreground/siren/69-Siren-1.wav' with backports.tempfile.TemporaryDirectory() as tmpdir: nonfile = os.path.join(tmpdir, 'notafile') source_file_list = [sourcefile, nonfile] pytest.raises(ScaperError, scaper.core._validate_source_file, ('choose', source_file_list), ('const', 'siren')) # must be const or choose bad_label_dists = [('uniform', 0, 1), ('normal', 0, 1), ('truncnorm', 0, 1, 0, 1)] for bld in bad_label_dists: pytest.raises(ScaperError, scaper.core._validate_source_file, bld, ('const', 'siren')) def test_validate_time(): def __test_bad_time_tuple(time_tuple): pytest.raises(ScaperError, scaper.core._validate_time, time_tuple) # bad consts bad_time_values = [None, -1, 1j, 'yes', [], [5]] for btv in bad_time_values: __test_bad_time_tuple(('const', btv)) # empty list for choose __test_bad_time_tuple(('choose', [])) # bad consts in list for choose for btv in bad_time_values: __test_bad_time_tuple(('choose', [btv])) # uniform can't have negative min value __test_bad_time_tuple(('uniform', -1, 1)) # using normal will issue a warning since it can generate neg values pytest.warns(ScaperWarning, scaper.core._validate_time, ('normal', 5, 2)) # truncnorm can't have negative min value __test_bad_time_tuple(('truncnorm', 0, 1, -1, 1)) def test_validate_duration(): def __test_bad_duration_tuple(duration_tuple): pytest.raises(ScaperError, scaper.core._validate_duration, duration_tuple) # bad consts bad_dur_values = [None, -1, 0, 1j, 'yes', [], [5]] for bdv in bad_dur_values: __test_bad_duration_tuple(('const', bdv)) # empty list for choose __test_bad_duration_tuple(('choose', [])) # bad consts in list for choose for bdv in bad_dur_values: __test_bad_duration_tuple(('choose', [bdv])) # uniform can't have negative or 0 min value __test_bad_duration_tuple(('uniform', -1, 1)) __test_bad_duration_tuple(('uniform', 0, 1)) # using normal will issue a warning since it can generate neg values pytest.warns(ScaperWarning, scaper.core._validate_duration, ('normal', 5, 2)) # truncnorm can't have negative or zero min value __test_bad_duration_tuple(('truncnorm', 0, 1, -1, 1)) __test_bad_duration_tuple(('truncnorm', 0, 1, 0, 1)) def test_validate_snr(): def __test_bad_snr_tuple(snr_tuple): pytest.raises(ScaperError, scaper.core._validate_snr, snr_tuple) # bad consts bad_snr_values = [None, 1j, 'yes', [], [5]] for bsv in bad_snr_values: __test_bad_snr_tuple(('const', bsv)) # empty list for choose __test_bad_snr_tuple(('choose', [])) # bad consts in list for choose for bsv in bad_snr_values: __test_bad_snr_tuple(('choose', [bsv])) def test_validate_pitch_shift(): def __test_bad_ps_tuple(ps_tuple): pytest.raises(ScaperError, scaper.core._validate_pitch_shift, ps_tuple) # bad consts bad_ps_values = [None, 1j, 'yes', [], [5]] for bv in bad_ps_values: __test_bad_ps_tuple(('const', bv)) # empty list for choose __test_bad_ps_tuple(('choose', [])) # bad consts in list for choose for bv in bad_ps_values: __test_bad_ps_tuple(('choose', [bv])) def test_validate_time_stretch(): def __test_bad_ts_tuple(ts_tuple): pytest.raises(ScaperError, scaper.core._validate_time_stretch, ts_tuple) # bad consts bad_ps_values = [None, 1j, 'yes', [], [5], -5, 0] for bv in bad_ps_values: __test_bad_ts_tuple(('const', bv)) # empty list for choose __test_bad_ts_tuple(('choose', [])) # bad consts in list for choose for bv in bad_ps_values: __test_bad_ts_tuple(('choose', [bv])) # bad start time in distributions __test_bad_ts_tuple(('uniform', 0, 1)) __test_bad_ts_tuple(('uniform', -5, 1)) __test_bad_ts_tuple(('truncnorm', 5, 1, 0, 10)) __test_bad_ts_tuple(('truncnorm', 5, 1, -5, 10)) # Using normal dist must raise warning since can give neg or 0 values pytest.warns( ScaperWarning, scaper.core._validate_time_stretch, ('normal', 5, 1)) def test_validate_event(): bad_allowed_labels = [0, 'yes', 1j, np.array([1, 2, 3])] for bal in bad_allowed_labels: pytest.raises(ScaperError, scaper.core._validate_event, label=('choose', []), source_file=('choose', []), source_time=('const', 0), event_time=('const', 0), event_duration=('const', 1), snr=('const', 0), allowed_labels=bal, pitch_shift=None, time_stretch=None) def test_scaper_init(): ''' Test creation of Scaper object. ''' # bad duration sc = pytest.raises(ScaperError, scaper.Scaper, -5, FG_PATH, BG_PATH) # all args valid sc = scaper.Scaper(10.0, FG_PATH, BG_PATH) assert sc.fg_path == FG_PATH assert sc.bg_path == BG_PATH # bad fg path sc = pytest.raises(ScaperError, scaper.Scaper, 10.0, 'tests/data/audio/wrong', BG_PATH) # bad bg path sc = pytest.raises(ScaperError, scaper.Scaper, 10.0, FG_PATH, 'tests/data/audio/wrong') # ensure fg_labels and bg_labels populated properly sc = scaper.Scaper(10.0, FG_PATH, BG_PATH) assert sc.fg_labels == FB_LABELS assert sc.bg_labels == BG_LABELS # ensure default values have been set assert sc.sr == 44100 assert sc.ref_db == -12 assert sc.n_channels == 1 assert sc.fade_in_len == 0.01 # 10 ms assert sc.fade_out_len == 0.01 # 10 ms def test_reset_fg_bg_event_spec(): def _add_fg_event(sc): sc.add_event(label=('const', 'siren'), source_file=('choose', []), source_time=('const', 0), event_time=('uniform', 0, 9), event_duration=('truncnorm', 2, 1, 1, 3), snr=('uniform', 10, 20), pitch_shift=('normal', 0, 1), time_stretch=('uniform', 0.8, 1.2)) def _add_bg_event(sc): sc.add_background(("const", "park"), ("choose", []), ("const", 0)) sc = scaper.Scaper( 10.0, fg_path=FG_PATH, bg_path=BG_PATH, random_state=0) # there should be no events initially assert len(sc.fg_spec) == 0 assert len(sc.bg_spec) == 0 # there should be one foreground event now _add_fg_event(sc) assert len(sc.fg_spec) == 1 first_fg_spec = deepcopy(sc.fg_spec) # after this there should be no foreground events sc.reset_fg_event_spec() assert len(sc.fg_spec) == 0 # add the foreground event back. now the original fg_spec and this one should be # the same. _add_fg_event(sc) assert first_fg_spec == sc.fg_spec # start over, this time using reset_bg_spec too. sc.reset_fg_event_spec() # there should be one background event and one foreground event now _add_fg_event(sc) _add_bg_event(sc) assert len(sc.fg_spec) == 1 assert len(sc.bg_spec) == 1 first_fg_spec = deepcopy(sc.fg_spec) first_bg_spec = deepcopy(sc.bg_spec) # after this there should be no foreground or background events sc.reset_fg_event_spec() sc.reset_bg_event_spec() assert len(sc.fg_spec) == 0 assert len(sc.bg_spec) == 0 # add the both events back. now both event spec sshould match the original _add_fg_event(sc) _add_bg_event(sc) assert first_fg_spec == sc.fg_spec assert first_bg_spec == sc.bg_spec def test_scaper_add_background(): ''' Test Scaper.add_background function ''' sc = scaper.Scaper(10.0, FG_PATH, BG_PATH) # Set concrete background label # label, source_file, source_time sc.add_background(("const", "park"), ("choose", []), ("const", 0)) # Check that event has been added to the background spec, and that the # values that are set automatically by this method (event_time, # event_duration, snr and role) are correctly set to their expected values. bg_event_expected = EventSpec(label=("const", "park"), source_file=("choose", []), source_time=("const", 0), event_time=("const", 0), event_duration=("const", sc.duration), snr=("const", 0), role='background', pitch_shift=None, time_stretch=None) assert sc.bg_spec == [bg_event_expected] def test_scaper_add_event(): sc = scaper.Scaper(10.0, FG_PATH, BG_PATH) # Initially fg_spec should be empty assert sc.fg_spec == [] # Add one event sc.add_event(label=('const', 'siren'), source_file=('choose', []), source_time=('const', 0), event_time=('uniform', 0, 9), event_duration=('truncnorm', 2, 1, 1, 3), snr=('uniform', 10, 20), pitch_shift=('normal', 0, 1), time_stretch=('uniform', 0.8, 1.2)) # Now should be one event in fg_spec assert len(sc.fg_spec) == 1 fg_event_expected = EventSpec(label=('const', 'siren'), source_file=('choose', []), source_time=('const', 0), event_time=('uniform', 0, 9), event_duration=('truncnorm', 2, 1, 1, 3), snr=('uniform', 10, 20), role='foreground', pitch_shift=('normal', 0, 1), time_stretch=('uniform', 0.8, 1.2)) assert sc.fg_spec[0] == fg_event_expected def test_scaper_instantiate_event(): # GF EVENT TO WORK WITH fg_event = EventSpec(label=('const', 'siren'), source_file=('choose', []), source_time=('const', 0), event_time=('uniform', 0, 9), event_duration=('truncnorm', 2, 1, 1, 3), snr=('uniform', 10, 20), role='foreground', pitch_shift=('normal', 0, 1), time_stretch=('uniform', 0.8, 1.2)) # test valid case sc = scaper.Scaper(10.0, fg_path=FG_PATH, bg_path=BG_PATH) instantiated_event = sc._instantiate_event( fg_event, isbackground=False, allow_repeated_label=True, allow_repeated_source=True, used_labels=[], used_source_files=[], disable_instantiation_warnings=True) assert instantiated_event.label == 'siren' assert instantiated_event.source_file == ( 'tests/data/audio/foreground/siren/69-Siren-1.wav') assert instantiated_event.source_time == 0 assert 0 <= instantiated_event.event_time <= 9 assert 1 <= instantiated_event.event_duration <= 3 assert 10 <= instantiated_event.snr <= 20 assert instantiated_event.role == 'foreground' assert scaper.util.is_real_number(instantiated_event.pitch_shift) assert 0.8 <= instantiated_event.time_stretch <= 1.2 # when a label needs to be replaced because it's used already fg_event8 = fg_event._replace(label=('choose', [])) # repeat several times to increase chance of hitting the line we need to # test for _ in range(20): instantiated_event = sc._instantiate_event( fg_event8, isbackground=False, allow_repeated_label=False, allow_repeated_source=True, used_labels=['siren', 'human_voice'], disable_instantiation_warnings=True) assert instantiated_event.label == 'car_horn' # when a source file needs to be replaced because it's used already fg_event9 = fg_event._replace(label=('const', 'human_voice')) # repeat several times to increase chance of hitting the line we need to # test for _ in range(20): instantiated_event = sc._instantiate_event( fg_event9, isbackground=False, allow_repeated_label=True, allow_repeated_source=False, used_labels=[], used_source_files=( ['tests/data/audio/foreground/human_voice/' '42-Human-Vocal-Voice-all-aboard_edit.wav', 'tests/data/audio/foreground/human_voice/' '42-Human-Vocal-Voice-taxi-1_edit.wav']), disable_instantiation_warnings=True) assert instantiated_event.source_file == ( 'tests/data/audio/foreground/human_voice/' '42-Human-Vocal-Voice-taxi-2_edit.wav') # Protected labels must have original source duration and source time 0 sc = scaper.Scaper(10.0, fg_path=FG_PATH, bg_path=BG_PATH, protected_labels='human_voice') fg_event10 = fg_event._replace( label=('const', 'human_voice'), source_file=('const', 'tests/data/audio/foreground/human_voice/' '42-Human-Vocal-Voice-taxi-2_edit.wav'), source_time=('const', 0.3), event_duration=('const', 0.4)) instantiated_event = sc._instantiate_event( fg_event10, disable_instantiation_warnings=True) assert instantiated_event.source_time == 0 assert np.allclose(instantiated_event.event_duration, 0.806236, atol=1e-5) # repeated label when not allowed throws error sc = scaper.Scaper(10.0, fg_path=FG_PATH, bg_path=BG_PATH) pytest.raises(ScaperError, sc._instantiate_event, fg_event, isbackground=False, allow_repeated_label=False, allow_repeated_source=True, used_labels=['siren']) # repeated source when not allowed throws error pytest.raises(ScaperError, sc._instantiate_event, fg_event, isbackground=False, allow_repeated_label=True, allow_repeated_source=False, used_labels=['siren'], used_source_files=( ['tests/data/audio/foreground/siren/69-Siren-1.wav'])) # event duration longer than source duration: warning fg_event2 = fg_event._replace(label=('const', 'car_horn'), event_duration=('const', 5)) pytest.warns(ScaperWarning, sc._instantiate_event, fg_event2) # event duration longer than soundscape duration: warning fg_event3 = fg_event._replace(event_time=('const', 0), event_duration=('const', 15), time_stretch=None) pytest.warns(ScaperWarning, sc._instantiate_event, fg_event3) # stretched event duration longer than soundscape duration: warning fg_event4 = fg_event._replace(event_time=('const', 0), event_duration=('const', 6), time_stretch=('const', 2)) pytest.warns(ScaperWarning, sc._instantiate_event, fg_event4) # 'const' source_time + event_duration > source_duration: warning fg_event5a = fg_event._replace(event_time=('const', 0), event_duration=('const', 8), source_time=('const', 20)) pytest.warns(ScaperWarning, sc._instantiate_event, fg_event5a) # 'choose' source_time + event_duration > source_duration: warning fg_event5b = fg_event._replace(event_time=('const', 0), event_duration=('const', 8), source_time=('choose', [20, 20])) pytest.warns(ScaperWarning, sc._instantiate_event, fg_event5b) # 'uniform' source_time + event_duration > source_duration: warning fg_event5c = fg_event._replace(event_time=('const', 0), event_duration=('const', 8), source_time=('uniform', 20, 25)) pytest.warns(ScaperWarning, sc._instantiate_event, fg_event5c) # 'normal' source_time + event_duration > source_duration: warning fg_event5d = fg_event._replace(event_time=('const', 0), event_duration=('const', 8), source_time=('normal', 20, 2)) pytest.warns(ScaperWarning, sc._instantiate_event, fg_event5d) # 'truncnorm' source_time + event_duration > source_duration: warning fg_event5e = fg_event._replace(event_time=('const', 0), event_duration=('const', 8), source_time=('truncnorm', 20, 2, 20, 20)) pytest.warns(ScaperWarning, sc._instantiate_event, fg_event5e) # 'normal' random draw above mean with mean = source_duration - event_duration # source_time + event_duration > source_duration: warning fg_event5f = fg_event._replace(event_time=('const', 0), event_duration=('const', 8), source_time=('normal', 18.25, 2)) def _repeat_instantiation(event): # keep going till we hit a draw that covers when the draw exceeds # source_duration - event_duration (18.25). Use max_draws # just in case so that testing is guaranteed to terminate. source_time = 0 num_draws = 0 max_draws = 1000 while source_time < 18.25 and num_draws < max_draws: instantiated_event = sc._instantiate_event(event) source_time = instantiated_event.source_time num_draws += 1 pytest.warns(ScaperWarning, _repeat_instantiation, fg_event5f) # event_time + event_duration > soundscape duration: warning fg_event6 = fg_event._replace(event_time=('const', 8), event_duration=('const', 5), time_stretch=None) pytest.warns(ScaperWarning, sc._instantiate_event, fg_event6) # event_time + stretched event_duration > soundscape duration: warning fg_event7 = fg_event._replace(event_time=('const', 5), event_duration=('const', 4), time_stretch=('const', 2)) pytest.warns(ScaperWarning, sc._instantiate_event, fg_event7) # stretched duration should always be adjusted to be <= self.duration for stretch in [2, 3, 1.5]: fg_event11 = fg_event._replace(event_time=('const', 2), event_duration=('const', 7), time_stretch=('const', stretch)) fg_event11_inst = sc._instantiate_event(fg_event11) assert fg_event11_inst.event_time == 0 assert fg_event11_inst.event_duration == sc.duration / stretch # Make sure event time is respected when possible for e_stretch, e_duration in zip([1, 1.25, 0.5], [7, 7, 18]): fg_event12 = fg_event._replace(event_time=('const', 1), event_duration=('const', e_duration), time_stretch=('const', e_stretch)) fg_event12_inst = sc._instantiate_event(fg_event12) assert fg_event12_inst.event_time == 1 assert fg_event12_inst.event_duration == e_duration def test_scaper_instantiate(): for sr in SAMPLE_RATES: REG_JAM_PATH = TEST_PATHS[sr]['REG'].jams # Here we just instantiate a known fixed spec and check if that jams # we get back is as expected. sc = scaper.Scaper(10.0, fg_path=FG_PATH, bg_path=BG_PATH) sc.ref_db = -50 sc.sr = sr # background sc.add_background( label=('const', 'park'), source_file=( 'const', 'tests/data/audio/background/park/' '268903__yonts__city-park-tel-aviv-israel.wav'), source_time=('const', 0)) # foreground events sc.add_event( label=('const', 'siren'), source_file=('const', 'tests/data/audio/foreground/' 'siren/69-Siren-1.wav'), source_time=('const', 5), event_time=('const', 2), event_duration=('const', 5), snr=('const', 5), pitch_shift=None, time_stretch=None) sc.add_event( label=('const', 'car_horn'), source_file=('const', 'tests/data/audio/foreground/' 'car_horn/17-CAR-Rolls-Royce-Horn.wav'), source_time=('const', 0), event_time=('const', 5), event_duration=('const', 2), snr=('const', 20), pitch_shift=('const', 1), time_stretch=None) sc.add_event( label=('const', 'human_voice'), source_file=('const', 'tests/data/audio/foreground/' 'human_voice/42-Human-Vocal-Voice-taxi-2_edit.wav'), source_time=('const', 0), event_time=('const', 7), event_duration=('const', 2), snr=('const', 10), pitch_shift=None, time_stretch=('const', 1.2)) # Instantiate jam = sc._instantiate(disable_instantiation_warnings=True) # Ignore all fields set by generate but not by instantiate sandbox_exclude = [ 'txt_path', 'jams_path', 'audio_path', 'no_audio', 'save_isolated_events', 'fix_clipping', 'peak_normalization', 'peak_normalization_scale_factor', 'quick_pitch_time', 'ref_db_change', 'ref_db_generated', 'txt_sep', 'disable_sox_warnings', 'disable_instantiation_warnings' ] # Load regression jam regjam = jams.load(REG_JAM_PATH) _compare_scaper_jams(jam, regjam, exclude_additional_scaper_sandbox_keys=sandbox_exclude) def test_generate_with_seeding(atol=1e-4, rtol=1e-8): # test a scaper generator with different random seeds. init with same random seed # over and over to make sure the output wav stays the same seeds = [ 0, 10, 20, scaper.util._check_random_state(0), scaper.util._check_random_state(10), scaper.util._check_random_state(20) ] num_generators = 2 for seed in seeds: generators = [] for i in range(num_generators): generators.append(_create_scaper_with_random_seed(deepcopy(seed))) _compare_generators(generators) def test_set_random_state(atol=1e-4, rtol=1e-8): # test a scaper generator with different random seeds. this time use # set_random_state to change the seed instead seeds = [ 0, 10, 20, scaper.util._check_random_state(0), scaper.util._check_random_state(10), scaper.util._check_random_state(20) ] num_generators = 2 for seed in seeds: generators = [] for i in range(num_generators): _sc = _create_scaper_with_random_seed(None) _sc.set_random_state(deepcopy(seed)) generators.append(_sc) _compare_generators(generators) def _compare_generators(generators, atol=1e-4, rtol=1e-8): tmpfiles = [] with _close_temp_files(tmpfiles): wav_files = [ tempfile.NamedTemporaryFile(suffix='.wav', delete=True) for i in range(len(generators)) ] jam_files = [ tempfile.NamedTemporaryFile(suffix='.jams', delete=True) for i in range(len(generators)) ] txt_files = [ tempfile.NamedTemporaryFile(suffix='.txt', delete=True) for i in range(len(generators)) ] tmpfiles += wav_files + jam_files + txt_files for i, sc in enumerate(generators): generators[i].generate( wav_files[i].name, jam_files[i].name, txt_path=txt_files[i].name, disable_instantiation_warnings=True ) audio = [soundfile.read(wav_file.name)[0] for wav_file in wav_files] for i, a in enumerate(audio): assert np.allclose(audio[0], a, atol=atol, rtol=rtol) # load all the jams data # make sure they are all the same as the first one exclude_sandbox = ['audio_path', 'jams_path', 'txt_path'] jams_data = [jams.load(jam_file.name) for jam_file in jam_files] for x in jams_data: _compare_scaper_jams( x, jams_data[0], exclude_additional_scaper_sandbox_keys=exclude_sandbox) # load the txt files and compare them def _load_txt(txt_file): txt_data = [] with open(txt_file.name) as file: reader = csv.reader(file, delimiter='\t') for row in reader: txt_data.append(row) txt_data = np.asarray(txt_data) return txt_data txt_data = [_load_txt(txt_file) for txt_file in txt_files] regtxt_data = txt_data[0] for t in txt_data: assert np.allclose([float(x) for x in t[:, 0]], [float(x) for x in regtxt_data[:, 0]]) assert np.allclose([float(x) for x in t[:, 1]], [float(x) for x in regtxt_data[:, 1]]) # compare labels assert (t[:, 2] == regtxt_data[:, 2]).all() def _create_scaper_with_random_seed(seed): sc = scaper.Scaper(10.0, fg_path=FG_PATH, bg_path=BG_PATH, random_state=deepcopy(seed)) sc.ref_db = -50 sc.sr = 44100 # background sc.add_background( label=('choose', []), source_file=('choose', []), source_time=('const', 0)) # foreground events sc.add_event( label=('choose', []), source_file=('choose', []), source_time=('uniform', 0, 8), event_time=('truncnorm', 4, 1, 0, 8), event_duration=('normal', 4, 1), snr=('const', 5), pitch_shift=None, time_stretch=None) sc.add_event( label=('choose', []), source_file=('choose', []), source_time=('uniform', 0, 8), event_time=('truncnorm', 4, 1, 0, 8), event_duration=('normal', 4, 1), snr=('const', 20), pitch_shift=None, time_stretch=None) sc.add_event( label=('choose', []), source_file=('choose', []), source_time=('const', 0), event_time=('const', 7), event_duration=('const', 2), snr=('const', 10), pitch_shift=None, time_stretch=None) return sc def test_generate_audio(): for sr in SAMPLE_RATES: for n_ch in range(1, 3): REG_WAV_PATH = TEST_PATHS[sr]['REG'].wav REG_BGONLY_WAV_PATH = TEST_PATHS[sr]['REG_BGONLY'].wav REG_REVERB_WAV_PATH = TEST_PATHS[sr]['REG_REVERB'].wav _test_generate_audio(sr, REG_WAV_PATH, REG_BGONLY_WAV_PATH, REG_REVERB_WAV_PATH, n_ch) def _test_generate_audio(SR, REG_WAV_PATH, REG_BGONLY_WAV_PATH, REG_REVERB_WAV_PATH, N_CHANNELS, atol=1e-4, rtol=1e-8): # Regression test: same spec, same audio (not this will fail if we update # any of the audio processing techniques used (e.g. change time stretching # algorithm. sc = scaper.Scaper(10.0, fg_path=FG_PATH, bg_path=BG_PATH) sc.ref_db = -50 sc.sr = SR sc.n_channels = N_CHANNELS print("TEST SR: {}, # OF CHANNELS: {}".format(SR, N_CHANNELS)) # background sc.add_background( label=('const', 'park'), source_file=( 'const', 'tests/data/audio/background/park/' '268903__yonts__city-park-tel-aviv-israel.wav'), source_time=('const', 0)) # foreground events sc.add_event( label=('const', 'siren'), source_file=('const', 'tests/data/audio/foreground/' 'siren/69-Siren-1.wav'), source_time=('const', 5), event_time=('const', 2), event_duration=('const', 5), snr=('const', 5), pitch_shift=None, time_stretch=None) sc.add_event( label=('const', 'car_horn'), source_file=('const', 'tests/data/audio/foreground/' 'car_horn/17-CAR-Rolls-Royce-Horn.wav'), source_time=('const', 0), event_time=('const', 5), event_duration=('const', 2), snr=('const', 20), pitch_shift=('const', 1), time_stretch=None) sc.add_event( label=('const', 'human_voice'), source_file=('const', 'tests/data/audio/foreground/' 'human_voice/42-Human-Vocal-Voice-taxi-2_edit.wav'), source_time=('const', 0), event_time=('const', 7), event_duration=('const', 2), snr=('const', 10), pitch_shift=None, time_stretch=('const', 1.2)) tmpfiles = [] with _close_temp_files(tmpfiles): wav_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) tmpfiles.append(wav_file) jam = sc._instantiate(disable_instantiation_warnings=True) sc._generate_audio(wav_file.name, jam.annotations[0]) # validate audio wav, sr = soundfile.read(wav_file.name, always_2d=True) regwav, sr = soundfile.read(REG_WAV_PATH, always_2d=True) # TODO: Add multi-channel regression data. if N_CHANNELS == 1: assert np.allclose(wav, regwav, atol=atol, rtol=rtol) # with reverb sc._generate_audio(wav_file.name, jam.annotations[0], reverb=0.2) # validate audio wav, sr = soundfile.read(wav_file.name, always_2d=True) regwav, sr = soundfile.read(REG_REVERB_WAV_PATH, always_2d=True) # TODO: Add multi-channel regression data. if N_CHANNELS == 1: assert np.allclose(wav, regwav, atol=atol, rtol=rtol) # Don't disable sox warnings (just to cover line) sc._generate_audio(wav_file.name, jam.annotations[0], disable_sox_warnings=False) # validate audio wav, sr = soundfile.read(wav_file.name, always_2d=True) regwav, sr = soundfile.read(REG_WAV_PATH, always_2d=True) # TODO: Add multi-channel regression data. if N_CHANNELS == 1: assert np.allclose(wav, regwav, atol=atol, rtol=rtol) # namespace must be scaper jam.annotations[0].namespace = 'tag_open' pytest.raises(ScaperError, sc._generate_audio, wav_file.name, jam.annotations[0]) # unsupported event role must raise error jam.annotations[0].namespace = 'scaper' jam.annotations[0].data[3].value['role'] = 'ewok' pytest.raises(ScaperError, sc._generate_audio, wav_file.name, jam.annotations[0]) # soundscape with no events will raise warning and won't generate audio sc = scaper.Scaper(10.0, fg_path=FG_PATH, bg_path=BG_PATH) sc.ref_db = -50 jam = sc._instantiate(disable_instantiation_warnings=True) pytest.warns(ScaperWarning, sc._generate_audio, wav_file.name, jam.annotations[0]) # soundscape with only one event will use transformer (regression test) sc = scaper.Scaper(10.0, fg_path=FG_PATH, bg_path=BG_PATH) sc.ref_db = -20 sc.sr = SR # background sc.add_background( label=('const', 'park'), source_file=('const', 'tests/data/audio/background/park/' '268903__yonts__city-park-tel-aviv-israel.wav'), source_time=('const', 0)) reverb = 0.2 jam = sc._instantiate(disable_instantiation_warnings=True, reverb=reverb) sc._generate_audio(wav_file.name, jam.annotations[0], reverb=reverb) # validate audio wav, sr = soundfile.read(wav_file.name, always_2d=True) regwav, sr = soundfile.read(REG_BGONLY_WAV_PATH, always_2d=True) # TODO: Add multi-channel regression data. if N_CHANNELS == 1: assert np.allclose(wav, regwav, atol=atol, rtol=rtol) def create_scaper_scene_without_random_seed(): sc = scaper.Scaper(10.0, fg_path=FG_PATH, bg_path=BG_PATH) sc.ref_db = -50 sc.sr = 44100 # background sc.add_background( label=('const', 'park'), source_file=( 'const', 'tests/data/audio/background/park/' '268903__yonts__city-park-tel-aviv-israel.wav'), source_time=('const', 0)) # foreground events sc.add_event( label=('const', 'siren'), source_file=('const', 'tests/data/audio/foreground/' 'siren/69-Siren-1.wav'), source_time=('uniform', 0, 5), event_time=('normal', 5, 1), event_duration=('truncnorm', 5, 1, 4, 6), snr=('const', 5), pitch_shift=None, time_stretch=None) sc.add_event( label=('const', 'car_horn'), source_file=('const', 'tests/data/audio/foreground/' 'car_horn/17-CAR-Rolls-Royce-Horn.wav'), source_time=('const', 0), event_time=('const', 5), event_duration=('truncnorm', 3, 1, 1, 10), snr=('uniform', 10, 20), pitch_shift=('uniform', -1, 1), time_stretch=(None)) sc.add_event( label=('const', 'human_voice'), source_file=('const', 'tests/data/audio/foreground/' 'human_voice/42-Human-Vocal-Voice-taxi-2_edit.wav'), source_time=('const', 0), event_time=('const', 7), event_duration=('const', 2), snr=('const', 10), pitch_shift=('uniform', -1, 1), time_stretch=('uniform', .8, 1.2)) return sc def _test_generate_isolated_events(SR, isolated_events_path=None, atol=1e-4, rtol=1e-8): sc = create_scaper_scene_without_random_seed() tmpfiles = [] @contextmanager def _delete_files(mix_file, directory): yield try: shutil.rmtree(directory) os.remove(mix_file) except: pass wav_file = 'tests/mix.wav' if isolated_events_path is None: isolated_events_path = 'tests/mix_events' with _delete_files(wav_file, isolated_events_path): jam = sc._instantiate(disable_instantiation_warnings=True) sc._generate_audio(wav_file, jam.annotations[0], save_isolated_events=True, isolated_events_path=isolated_events_path) source_directory = os.path.splitext(wav_file)[0] + '_events' isolated_events = [] ann = jam.annotations.search(namespace='scaper')[0] soundscape_audio, _ = soundfile.read(ann.sandbox.scaper.soundscape_audio_path) isolated_event_audio_paths = ann.sandbox.scaper.isolated_events_audio_path isolated_audio = [] role_counter = { 'background': 0, 'foreground': 0 } for event_spec, event_audio_path in zip(ann, isolated_event_audio_paths): # event_spec contains the event description, label, etc # event_audio contains the path to the actual audio # make sure the path matches the event description look_for = '{:s}{:d}_{:s}.wav'.format( event_spec.value['role'], role_counter[event_spec.value['role']], event_spec.value['label'] ) expected_path = os.path.join(isolated_events_path, look_for) # make sure the path exists assert os.path.exists(expected_path) # make sure what's in the sandbox also exists assert os.path.exists(event_audio_path) # is an audio file with the same contents as what we expect _isolated_expected_audio, sr = soundfile.read(expected_path) _isolated_sandbox_audio, sr = soundfile.read(event_audio_path) assert np.allclose(_isolated_sandbox_audio, _isolated_expected_audio) # make sure the filename matches assert look_for == os.path.basename(event_audio_path) # increment for the next role role_counter[event_spec.value['role']] += 1 isolated_audio.append(_isolated_sandbox_audio) # the sum of the isolated audio should sum to the soundscape assert np.allclose(sum(isolated_audio), soundscape_audio, atol=1e-8, rtol=1e-8) jam = sc._instantiate(disable_instantiation_warnings=True) # Running with save_isolated_events = True and reverb not None raises a warning pytest.warns(ScaperWarning, sc._generate_audio, wav_file, jam.annotations[0], save_isolated_events=True, reverb=.5) def test_generate_isolated_events(): for sr, isolated_events_path in zip( (16000, 22050, 44100), (None, 'tests/mix_events', None)): # try it a bunch of times for i in range(10): _test_generate_isolated_events(sr, isolated_events_path) def test_generate(): for sr in SAMPLE_RATES: REG_WAV_PATH, REG_JAM_PATH, REG_TXT_PATH = TEST_PATHS[sr]['REG'] _test_generate(sr, REG_WAV_PATH, REG_JAM_PATH, REG_TXT_PATH) def _make_test_generate_scaper(SR): sc = scaper.Scaper(10.0, fg_path=FG_PATH, bg_path=BG_PATH) sc.ref_db = -50 sc.sr = SR # background sc.add_background( label=('const', 'park'), source_file=( 'const', 'tests/data/audio/background/park/' '268903__yonts__city-park-tel-aviv-israel.wav'), source_time=('const', 0)) # foreground events sc.add_event( label=('const', 'siren'), source_file=('const', 'tests/data/audio/foreground/' 'siren/69-Siren-1.wav'), source_time=('const', 5), event_time=('const', 2), event_duration=('const', 5), snr=('const', 5), pitch_shift=None, time_stretch=None) sc.add_event( label=('const', 'car_horn'), source_file=('const', 'tests/data/audio/foreground/' 'car_horn/17-CAR-Rolls-Royce-Horn.wav'), source_time=('const', 0), event_time=('const', 5), event_duration=('const', 2), snr=('const', 20), pitch_shift=('const', 1), time_stretch=None) sc.add_event( label=('const', 'human_voice'), source_file=('const', 'tests/data/audio/foreground/' 'human_voice/42-Human-Vocal-Voice-taxi-2_edit.wav'), source_time=('const', 0), event_time=('const', 7), event_duration=('const', 2), snr=('const', 10), pitch_shift=None, time_stretch=('const', 1.2)) return sc def _test_generate(SR, REG_WAV_PATH, REG_JAM_PATH, REG_TXT_PATH, atol=1e-4, rtol=1e-8): sc = _make_test_generate_scaper(SR) tmpfiles = [] with _close_temp_files(tmpfiles): wav_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) jam_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True) txt_file = tempfile.NamedTemporaryFile(suffix='.txt', delete=True) tmpfiles.append(wav_file) tmpfiles.append(jam_file) tmpfiles.append(txt_file) sc.generate(wav_file.name, jam_file.name, txt_path=txt_file.name, disable_instantiation_warnings=True) # validate audio wav, sr = soundfile.read(wav_file.name) regwav, sr = soundfile.read(REG_WAV_PATH) assert np.allclose(wav, regwav, atol=atol, rtol=rtol) # validate jams jam = jams.load(jam_file.name) regjam = jams.load(REG_JAM_PATH) sandbox_exclude = ['audio_path', 'jams_path', 'txt_path'] _compare_scaper_jams( jam, regjam, exclude_additional_scaper_sandbox_keys=sandbox_exclude) # validate txt _compare_txt_annotation(txt_file.name, REG_TXT_PATH) # reverb value must be in (0, 1) range for reverb in [-1, 2]: pytest.raises(ScaperError, sc.generate, wav_file.name, jam_file.name, reverb=reverb, disable_instantiation_warnings=True) def test_generate_return_api(): for sr in SAMPLE_RATES: REG_WAV_PATH, REG_JAM_PATH, REG_TXT_PATH = TEST_PATHS[sr]['REG'] _test_generate_return_api(sr, REG_WAV_PATH, REG_JAM_PATH, REG_TXT_PATH) def _test_generate_return_api(SR, REG_WAV_PATH, REG_JAM_PATH, REG_TXT_PATH, atol=1e-4, rtol=1e-8): sc = _make_test_generate_scaper(SR) tmpfiles = [] with _close_temp_files(tmpfiles): wav_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) jam_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True) txt_file = tempfile.NamedTemporaryFile(suffix='.txt', delete=True) tmpfiles.append(wav_file) tmpfiles.append(jam_file) tmpfiles.append(txt_file) with backports.tempfile.TemporaryDirectory() as isolated_events_path: saved_events_path = os.path.join(isolated_events_path, 'original') (soundscape_audio, soundscape_jam, annotation_list, event_audio_list) = \ sc.generate(audio_path=wav_file.name, jams_path=jam_file.name, txt_path=txt_file.name, save_isolated_events=True, isolated_events_path=saved_events_path, disable_instantiation_warnings=True) # print(soundscape_audio.shape) # validate audio wav, sr = soundfile.read(wav_file.name, always_2d=True) regwav, sr = soundfile.read(REG_WAV_PATH, always_2d=True) # print(wav.shape) # print(regwav.shape) # assert 1 == 2 assert np.allclose(soundscape_audio, wav, atol=atol, rtol=rtol) assert np.allclose(soundscape_audio, regwav, atol=atol, rtol=rtol) # validate jams jam = jams.load(jam_file.name) regjam = jams.load(REG_JAM_PATH) sandbox_exclude = [ 'isolated_events_path', 'save_isolated_events', 'jams_path', 'txt_path', 'audio_path'] _compare_scaper_jams(soundscape_jam, jam) _compare_scaper_jams( soundscape_jam, regjam, exclude_additional_scaper_sandbox_keys=sandbox_exclude) # validate txt annotation txt_data = _load_txt_annotation(txt_file.name) reg_txt_data = _load_txt_annotation(REG_TXT_PATH) _compare_txt_annotation(annotation_list, txt_data) _compare_txt_annotation(annotation_list, reg_txt_data) # validate event audio saved_event_files = [ os.path.join(saved_events_path, x) for x in sorted(os.listdir(saved_events_path)) ] orig_event_audio = [soundfile.read(x, always_2d=True)[0] for x in saved_event_files] for return_event, orig_event in zip(event_audio_list, orig_event_audio): assert np.allclose(return_event, orig_event, atol=1e-8, rtol=rtol) def test_scaper_off_by_one_with_jams(): # this file broke in Scaper 1.3.3 and before as the duration # of the generated audio was incorrect. it was addressed by PR #88. # using it to test if it will ever break again jam_file = 'tests/data/regression/scaper_133_off_by_one_regression_test.jams' tmpfiles = [] with _close_temp_files(tmpfiles): gen_wav_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) scaper.generate_from_jams(jam_file, gen_wav_file.name) gen_wav, sr = soundfile.read(gen_wav_file.name) assert gen_wav.shape[0] == 10 * 44100 def test_backwards_compat_for_duration(): for sr in SAMPLE_RATES: REG_JAM_PATH = TEST_PATHS[sr]['REG'].jams tmpfiles = [] with _close_temp_files(tmpfiles): orig_wav = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) gen_wav = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) jam_without_orig_duration = tempfile.NamedTemporaryFile( suffix='.jams', delete=True) jam = jams.load(REG_JAM_PATH) scaper.generate_from_jams(REG_JAM_PATH, orig_wav.name) ann = jam.annotations[0] ann.sandbox.scaper.pop('original_duration') jam.save(jam_without_orig_duration.name) scaper.generate_from_jams( jam_without_orig_duration.name, gen_wav.name) orig_audio, sr = soundfile.read(orig_wav.name) gen_audio, sr = soundfile.read(gen_wav.name) assert np.allclose(orig_audio, gen_audio) pytest.warns(ScaperWarning, scaper.generate_from_jams, jam_without_orig_duration.name, gen_wav.name) def _generate_soundscape_with_short_background(background_file, audio_path, jams_path, ref_db): with backports.tempfile.TemporaryDirectory() as tmpdir: subdir = os.path.join(tmpdir, 'audio') shutil.copytree(SHORT_BG_PATH, subdir) OUTPUT_PATH = os.path.join(subdir, 'noise', 'noise.wav') shutil.copyfile(background_file, OUTPUT_PATH) sc = scaper.Scaper(10, FG_PATH, subdir, random_state=0) sc.sr = 16000 sc.ref_db = ref_db sc.fade_in_len = 0 sc.fade_out_len = 0 sc.add_background( label=('const', 'noise'), source_file=('const', OUTPUT_PATH), source_time=('const', 0) ) sc.generate(audio_path, jams_path) def test_scaper_generate_with_fade(): # Test scaper generate with different fade lengths # Works by using a fade of 0 at first then comparing # samples of the event using different fades. fade_lens = [0, 0.01, 0.05, 0.1] outputs = {} for fade_in in fade_lens: for fade_out in fade_lens: sc = scaper.Scaper(0.2, FG_PATH, BG_PATH, random_state=0) sc.sr = 16000 sc.ref_db = -20 sc.fade_in_len = fade_in sc.fade_out_len = fade_out sc.add_event( label=('const', 'siren'), source_file=('choose', []), source_time=('uniform', 0, 10), event_time=('const', 0), event_duration=('const', 0.2), snr=('uniform', -5, 5), pitch_shift=('uniform', -1, 1), time_stretch=('uniform', 0.8, 1.2)) _, _, _, event_audio_list = sc.generate() outputs[(fade_in, fade_out)] = event_audio_list[0] no_fade = outputs[(0, 0)] for key, val in outputs.items(): fade_in, fade_out = key fade_in_samples, fade_out_samples = ( int(fade_in * sc.sr), int(fade_out * sc.sr) ) # Compare first fade_in_samples with no_fade if fade_in_samples > 0: ratio = val[:fade_in_samples] / no_fade[:fade_in_samples] fade_in_window = np.sin( np.linspace(0, np.pi / 2, fade_in_samples))[..., None] mask = np.invert(np.isnan(ratio)) assert np.allclose(ratio[mask], fade_in_window[mask]) if fade_out_samples > 0: ratio = val[-fade_out_samples:] / no_fade[-fade_out_samples:] fade_out_window = np.sin( np.linspace(np.pi / 2, 0, fade_out_samples))[..., None] # Ignore points where the signal has no energy mask = np.invert(np.isnan(ratio)) assert np.allclose(ratio[mask], fade_out_window[mask]) def test_scaper_with_short_background(): SHORT_BG_FILE = os.path.join( SHORT_BG_PATH, 'noise', 'noise-free-sound-0145.wav') tmpfiles = [] with _close_temp_files(tmpfiles): jam_file = tempfile.NamedTemporaryFile(suffix='.jams', delete=True) tiled_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) wav1_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) wav2_file = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) tmpfiles.append(wav1_file) tmpfiles.append(wav2_file) tmpfiles.append(tiled_file) tmpfiles.append(jam_file) _generate_soundscape_with_short_background( SHORT_BG_FILE, wav1_file.name, jam_file.name, ref_db=-40) # what it should be is the file tiled with itself and then cut to 10s # write it to disk and then use it in a new scaper object source_audio, sr = soundfile.read(SHORT_BG_FILE) duration_samples = int(10 * sr) # tile the audio to what we expect tiled_audio = np.tile( source_audio, 1 + int(duration_samples / source_audio.shape[0])) # cut it to what we want tiled_audio = tiled_audio[:duration_samples] # save it somewhere to be used in a new Scaper object soundfile.write(tiled_file.name, tiled_audio, sr) _generate_soundscape_with_short_background( tiled_file.name, wav2_file.name, jam_file.name, ref_db=-40) # compare what is generated with a short bg compared to a long bg # should be the same audio1, sr = soundfile.read(wav1_file.name) audio2, sr = soundfile.read(wav2_file.name) assert np.allclose(audio1, audio2) def test_clipping_and_normalization(): for sr in [16000, 44100]: sc = scaper.Scaper(10, FG_PATH, BG_PATH, random_state=0) sc.sr = sr sc.ref_db = -20 sc.add_event( label=('const', 'siren'), source_file=('choose', []), source_time=('uniform', 0, 10), event_time=('const', 0), event_duration=('const', 10), snr=('const', 20), pitch_shift=None, time_stretch=None) # extreme clipping sc_extreme = scaper.Scaper(10, FG_PATH, BG_PATH, random_state=0) sc_extreme.sr = 16000 sc_extreme.ref_db = -20 sc_extreme.add_event( label=('const', 'siren'), source_file=('choose', []), source_time=('uniform', 0, 10), event_time=('const', 0), event_duration=('const', 10), snr=('const', 40), pitch_shift=None, time_stretch=None) tmpfiles = [] with _close_temp_files(tmpfiles): # Make sure a warning is raised when there's clipping audio_path = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) tmpfiles.append(audio_path) pytest.warns(ScaperWarning, sc.generate, audio_path, fix_clipping=False) # Make sure a second warning is raised if we're fixing the clipping audio_path2 = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) tmpfiles.append(audio_path2) with pytest.warns(None) as record: sc.generate(audio_path2, fix_clipping=True) assert len(record) == 2 assert str(record[0].message) == 'Soundscape audio is clipping!' assert 'Peak normalization applied to fix clipping' in str(record[1].message) # Make sure we get a third warning when the scaling factor is < 0.05 audio_path3 = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) tmpfiles.append(audio_path3) with pytest.warns(None) as record: sc_extreme.generate(audio_path3, fix_clipping=True) assert len(record) == 3 assert str(record[0].message) == 'Soundscape audio is clipping!' assert 'Peak normalization applied to fix clipping' in str(record[1].message) assert 'Scale factor for peak normalization is extreme' in str(record[2].message) # PEAK NORMALIZATION TESTS # Make sure a warning is raised when there's clipping audio_path4 = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) tmpfiles.append(audio_path4) pytest.warns(ScaperWarning, sc.generate, audio_path4, fix_clipping=False, peak_normalization=True) # Make sure a second warning is NOT raised if we're peak normalizing by default audio_path5 = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) tmpfiles.append(audio_path5) with pytest.warns(None) as record: sc.generate(audio_path5, fix_clipping=False, peak_normalization=True) assert len(record) == 1 assert str(record[0].message) == 'Soundscape audio is clipping!' # Make sure we get two warnings when w're normalizing but not fixing # clipping explicitly and the scaling factor is < 0.05 audio_path6 = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) tmpfiles.append(audio_path6) with pytest.warns(None) as record: sc_extreme.generate(audio_path6, fix_clipping=False, peak_normalization=True) assert len(record) == 2 assert str(record[0].message) == 'Soundscape audio is clipping!' assert 'Scale factor for peak normalization is extreme' in str(record[1].message)
99,295
39.661753
119
py
scaper
scaper-master/tests/profile_speed.py
""" This is a profiling script to check the performance of Scaper. It generates 100 soundscapes in sequence (no parallelization). Running it on 2019 Macbook Pro currently takes 158.68 seconds (02:38). """ import scaper import numpy as np import tempfile import os import tqdm import zipfile import subprocess import time import csv import platform import psutil import datetime import math import multiprocessing import argparse import sys parser = argparse.ArgumentParser() parser.add_argument('--quick', action='store_true') args = parser.parse_args() cmd_line = ' '.join(sys.argv) cmd_line = 'python ' + cmd_line # Download the audio automatically FIX_DIR = 'tests/data/' QUICK_PITCH_TIME = args.quick def get_git_commit_hash(): process = subprocess.Popen( ['git', 'rev-parse', 'HEAD'], shell=False, stdout=subprocess.PIPE) git_head_hash = process.communicate()[0].strip().decode('utf-8') return git_head_hash def convert_size(size_bytes): if size_bytes == 0: return "0B" size_name = ("B", "KB", "MB", "GB", "TB", "PB", "EB", "ZB", "YB") i = int(math.floor(math.log(size_bytes, 1024))) p = math.pow(1024, i) s = round(size_bytes / p, 2) return "%s %s" % (s, size_name[i]) with tempfile.TemporaryDirectory() as tmpdir: path_to_audio = os.path.join(FIX_DIR, 'audio/') # OUTPUT FOLDER outfolder = tmpdir # SCAPER SETTINGS fg_folder = os.path.join(path_to_audio, 'foreground') bg_folder = os.path.join(path_to_audio, 'background') # If we parallelize this script, change this accordingly n_workers = 1 n_soundscapes = 100 ref_db = -50 duration = 10.0 min_events = 1 max_events = 9 event_time_dist = 'truncnorm' event_time_mean = 5.0 event_time_std = 2.0 event_time_min = 0.0 event_time_max = 10.0 source_time_dist = 'const' source_time = 0.0 event_duration_dist = 'uniform' event_duration_min = 0.5 event_duration_max = 4.0 snr_dist = 'uniform' snr_min = 6 snr_max = 30 pitch_dist = 'uniform' pitch_min = -3.0 pitch_max = 3.0 time_stretch_dist = 'uniform' time_stretch_min = 0.8 time_stretch_max = 1.2 # generate a random seed for this Scaper object seed = 123 # create a scaper that will be used below sc = scaper.Scaper(duration, fg_folder, bg_folder, random_state=seed) sc.protected_labels = [] sc.ref_db = ref_db # Generate 100 soundscapes using a truncated normal distribution of start times start_time = time.time() for n in tqdm.trange(n_soundscapes): print('Generating soundscape: {:d}/{:d}'.format(n+1, n_soundscapes)) # reset the event specifications for foreground and background at the # beginning of each loop to clear all previously added events sc.reset_bg_event_spec() sc.reset_fg_event_spec() # add background sc.add_background(label=('choose', []), source_file=('choose', []), source_time=('const', 0)) sc.fade_in_len = 0.01 sc.fade_out_len = 0.01 # add random number of foreground events n_events = np.random.randint(min_events, max_events+1) for _ in range(n_events): sc.add_event(label=('choose', []), source_file=('choose', []), source_time=(source_time_dist, source_time), event_time=(event_time_dist, event_time_mean, event_time_std, event_time_min, event_time_max), event_duration=(event_duration_dist, event_duration_min, event_duration_max), snr=(snr_dist, snr_min, snr_max), pitch_shift=(pitch_dist, pitch_min, pitch_max), time_stretch=(time_stretch_dist, time_stretch_min, time_stretch_max) ) # generate audiofile = os.path.join(outfolder, "soundscape_unimodal{:d}.wav".format(n)) jamsfile = os.path.join(outfolder, "soundscape_unimodal{:d}.jams".format(n)) txtfile = os.path.join(outfolder, "soundscape_unimodal{:d}.txt".format(n)) sc.generate(audiofile, jamsfile, allow_repeated_label=True, allow_repeated_source=True, reverb=0.1, disable_sox_warnings=True, quick_pitch_time=QUICK_PITCH_TIME, no_audio=False, txt_path=txtfile) time_taken = time.time() - start_time uname = platform.uname() row = { 'command': cmd_line, 'time_of_run': str(datetime.datetime.now()), 'scaper_version': scaper.__version__, 'python_version': platform.python_version(), 'system': uname.system, 'machine': uname.machine, 'processor': uname.processor, 'n_cpu': multiprocessing.cpu_count(), 'n_workers': n_workers, 'memory': convert_size(psutil.virtual_memory().total), 'n_soundscapes': n_soundscapes, 'execution_time': np.round(time_taken, 4), 'git_commit_hash': get_git_commit_hash(), } fieldnames = list(row.keys()) results_path = 'tests/profile_results.csv' write_header = not os.path.exists(results_path) with open(results_path, 'a') as f: writer = csv.DictWriter(f, fieldnames=fieldnames) if write_header: writer.writeheader() writer.writerow(row) with open(results_path, 'r') as f: csv_f = csv.reader(f) for row in csv_f: print('{:<30} {:<15} {:<15} {:<10} {:<10} {:<10} {:<5} {:<10} {:<10} {:<15} {:<10} {:}'.format(*row))
5,749
30.944444
122
py
scaper
scaper-master/tests/test_util.py
# CREATED: 10/15/16 7:52 PM by Justin Salamon <justin.salamon@nyu.edu> ''' Tests for functions in util.py ''' from scaper.util import _close_temp_files from scaper.util import _set_temp_logging_level from scaper.util import _validate_folder_path from scaper.util import _get_sorted_files from scaper.util import _populate_label_list from scaper.util import _sample_trunc_norm, _sample_choose, _sample_choose_weighted from scaper.util import max_polyphony from scaper.util import polyphony_gini from scaper.util import is_real_number, is_real_array from scaper.util import _check_random_state from scaper.scaper_exceptions import ScaperError from scaper.scaper_warnings import ScaperWarning import tempfile import os import logging import pytest import shutil import numpy as np from scipy.stats import truncnorm import jams from scaper.core import EventSpec from scaper import Scaper # FIXTURES BG_PATH = 'tests/data/audio/background/' FG_PATH = 'tests/data/audio/foreground/' FG_PATH_HUMANVOICE = 'tests/data/audio/foreground/human_voice' FG_LABEL_LIST = ['car_horn', 'human_voice', 'siren'] HUMANVOICE_FILES = ( [os.path.join(FG_PATH_HUMANVOICE, '42-Human-Vocal-Voice-all-aboard_edit.wav'), os.path.join(FG_PATH_HUMANVOICE, '42-Human-Vocal-Voice-taxi-1_edit.wav'), os.path.join(FG_PATH_HUMANVOICE, '42-Human-Vocal-Voice-taxi-2_edit.wav')]) SIREN_FILE = os.path.join(FG_PATH, 'siren', '69-Siren-1.wav') def test_close_temp_files(): ''' Create a bunch of temp files and then make sure they've been closed and deleted. ''' # With delete=True tmpfiles = [] with _close_temp_files(tmpfiles): for _ in range(5): tmpfiles.append( tempfile.NamedTemporaryFile(suffix='.wav', delete=True)) for tf in tmpfiles: assert tf.file.closed assert not os.path.isfile(tf.name) # With delete=False tmpfiles = [] with _close_temp_files(tmpfiles): for _ in range(5): tmpfiles.append( tempfile.NamedTemporaryFile(suffix='.wav', delete=False)) for tf in tmpfiles: assert tf.file.closed assert not os.path.isfile(tf.name) # with an exception before exiting try: tmpfiles = [] with _close_temp_files(tmpfiles): tmpfiles.append( tempfile.NamedTemporaryFile(suffix='.wav', delete=True)) raise ScaperError except ScaperError: for tf in tmpfiles: assert tf.file.closed assert not os.path.isfile(tf.name) else: assert False, 'Exception was not reraised.' def test_set_temp_logging_level(): ''' Ensure temp logging level is set as expected ''' logger = logging.getLogger() logger.setLevel('DEBUG') with _set_temp_logging_level('CRITICAL'): assert logging.getLevelName(logger.level) == 'CRITICAL' assert logging.getLevelName(logger.level) == 'DEBUG' def test_get_sorted_files(): ''' Ensure files are returned in expected order. ''' assert _get_sorted_files(FG_PATH_HUMANVOICE) == HUMANVOICE_FILES def test_validate_folder_path(): ''' Make sure invalid folder paths are caught ''' # bad folder path pytest.raises(ScaperError, _validate_folder_path, '/path/to/invalid/folder/') # good folder path should raise no error # make temp folder tmpdir = tempfile.mkdtemp() # validate _validate_folder_path(tmpdir) # remove it shutil.rmtree(tmpdir) def test_populate_label_list(): ''' Should add folder names contained within provided folder to provided list. ''' labellist = [] _populate_label_list(FG_PATH, labellist) assert sorted(labellist) == sorted(FG_LABEL_LIST) def test_check_random_state(): # seed is None rng_type = type(np.random.RandomState(10)) rng = _check_random_state(None) assert type(rng) == rng_type # seed is int rng = _check_random_state(10) assert type(rng) == rng_type # seed is RandomState rng_test = np.random.RandomState(10) rng = _check_random_state(rng_test) assert type(rng) == rng_type # seed is none of the above : error pytest.raises(ValueError, _check_random_state, 'random') def test_sample_choose(): # using choose with duplicates will issue a warning rng = _check_random_state(0) pytest.warns(ScaperWarning, _sample_choose, [0, 1, 2, 2, 2], rng) def test_sample_choose_weighted(): # make sure probabilities are factored in rng = _check_random_state(0) assert _sample_choose_weighted([0, 1, 2], [1, 0, 0], rng) == 0 assert _sample_choose_weighted([0, 1, 2], [0, 1, 0], rng) == 1 assert _sample_choose_weighted([0, 1, 2], [0, 0, 1], rng) == 2 samples = [] for _ in range(100000): samples.append(_sample_choose_weighted([0, 1], [0.3, 0.7], rng)) samples = np.asarray(samples) zero_ratio = (samples == 0).sum() / len(samples) one_ratio = (samples == 1).sum() / len(samples) assert np.allclose(zero_ratio, 0.3, atol=1e-2) assert np.allclose(one_ratio, 0.7, atol=1e-2) def test_sample_trunc_norm(): ''' Should return values from a truncated normal distribution. ''' rng = _check_random_state(0) # sample values from a distribution mu, sigma, trunc_min, trunc_max = 2, 1, 0, 5 x = [_sample_trunc_norm(mu, sigma, trunc_min, trunc_max, random_state=rng) for _ in range(100000)] x = np.asarray(x) # simple check: values must be within truncated bounds assert (x >= trunc_min).all() and (x <= trunc_max).all() # trickier check: values must approximate distribution's PDF hist, bins = np.histogram(x, bins=np.arange(0, 10.1, 0.2), density=True) xticks = bins[:-1] + 0.1 a, b = (trunc_min - mu) / float(sigma), (trunc_max - mu) / float(sigma) trunc_closed = truncnorm.pdf(xticks, a, b, mu, sigma) assert np.allclose(hist, trunc_closed, atol=0.015) def test_max_polyphony(): ''' Test the computation of polyphony of a scaper soundscape instantiation. ''' def __create_annotation_with_overlapping_events(n_events): ann = jams.Annotation(namespace='scaper') ann.duration = n_events / 2. + 10 for ind in range(n_events): instantiated_event = EventSpec(label='siren', source_file='/the/source/file.wav', source_time=0, event_time=ind / 2., event_duration=10, snr=0, role='foreground', pitch_shift=None, time_stretch=None) ann.append(time=ind / 2., duration=10, value=instantiated_event._asdict(), confidence=1.0) return ann def __create_annotation_without_overlapping_events(n_events): ann = jams.Annotation(namespace='scaper') ann.duration = n_events * 10 for ind in range(n_events): instantiated_event = EventSpec(label='siren', source_file='/the/source/file.wav', source_time=0, event_time=ind * 10, event_duration=5, snr=0, role='foreground', pitch_shift=None, time_stretch=None) ann.append(time=ind * 10, duration=5, value=instantiated_event._asdict(), confidence=1.0) return ann # 0 through 10 overlapping events for poly in range(11): ann = __create_annotation_with_overlapping_events(poly) est_poly = max_polyphony(ann) assert est_poly == poly # 1 through 10 NON-overlapping events for n_events in range(1, 11): ann = __create_annotation_without_overlapping_events(n_events) est_poly = max_polyphony(ann) assert est_poly == 1 def test_polyphony_gini(): ''' Test computation of polyphony gini ''' # Annotation must have namespace scaper, otherwise raise error ann = jams.Annotation('tag_open', duration=10) gini = pytest.raises(ScaperError, polyphony_gini, ann) # Annotation without duration set should raise error ann = jams.Annotation('scaper', duration=None) gini = pytest.raises(ScaperError, polyphony_gini, ann) # Annotation with no foreground events returns a gini of 0 sc = Scaper(10.0, FG_PATH, BG_PATH) # add background sc.add_background(label=('choose', []), source_file=('choose', []), source_time=('const', 0)) jam = sc._instantiate() ann = jam.annotations[0] gini = polyphony_gini(ann) assert gini == 0 def __test_gini_from_event_times(event_time_list, expected_gini, hop_size=0.01): print(event_time_list) # create scaper sc = Scaper(10.0, FG_PATH, BG_PATH) # add background sc.add_background(label=('choose', []), source_file=('choose', []), source_time=('const', 0)) # add foreground events based on the event time list # always use siren file since it is 26 s long, so we can choose the # event duration flexibly for onset, offset in event_time_list: sc.add_event(label=('const', 'siren'), source_file=('const', SIREN_FILE), source_time=('const', 0), event_time=('const', onset), event_duration=('const', offset - onset), snr=('uniform', 6, 30), pitch_shift=('uniform', -3, 3), time_stretch=None) jam = sc._instantiate() ann = jam.annotations[0] gini = polyphony_gini(ann, hop_size=hop_size) print(gini, expected_gini) assert np.allclose([gini], [expected_gini], atol=1e-5) event_time_lists = ([ [], [(0, 1)], [(0, 5), (5, 10)], [(0, 10), (3, 7), (4, 6)] ]) expected_ginis = [0, 0.1, 1, 0.75] for etl, g in zip(event_time_lists, expected_ginis): __test_gini_from_event_times(etl, g, hop_size=0.01) for etl, g in zip(event_time_lists, expected_ginis): __test_gini_from_event_times(etl, g, hop_size=1.0) def test_is_real_number(): non_reals = [None, 1j, 'yes'] yes_reals = [-1e12, -1, -1.0, 0, 1, 1.0, 1e12] # test single value for nr in non_reals: assert not is_real_number(nr) for yr in yes_reals: assert is_real_number(yr) def test_is_real_array(): non_reals = [None, 1j, 'yes'] yes_reals = [-1e12, -1, -1.0, 0, 1, 1.0, 1e12] # non-list non-array types must return false for x in non_reals + yes_reals: assert not is_real_array(x) # test array for nr in non_reals: assert not is_real_array([nr]) for yr in yes_reals: assert is_real_array([yr])
11,535
30.605479
102
py
scaper
scaper-master/tests/create_regression_data.py
import os import scaper import jams os.chdir('..') # FIXTURES # Paths to files for testing FG_PATH = 'tests/data/audio/foreground' BG_PATH = 'tests/data/audio/background' ALT_FG_PATH = 'tests/data/audio_alt_path/foreground' ALT_BG_PATH = 'tests/data/audio_alt_path/background' REG_NAME = 'soundscape_20200501' # REG_NAME = 'soundscape_20190326_22050' # REG_WAV_PATH = 'tests/data/regression/soundscape_20170928.wav' # REG_JAM_PATH = 'tests/data/regression/soundscape_20170928.jams' # REG_TXT_PATH = 'tests/data/regression/soundscape_20170928.txt' REG_BGONLY_NAME = 'bgonly_soundscape_20200501' # REG_BGONLY_NAME = 'bgonly_soundscape_20190326_22050' # REG_BGONLY_WAV_PATH = 'tests/data/regression/bgonly_soundscape_20170928.wav' # REG_BGONLY_JAM_PATH = 'tests/data/regression/bgonly_soundscape_20170928.jams' # REG_BGONLY_TXT_PATH = 'tests/data/regression/bgonly_soundscape_20170928.txt' REG_REVERB_NAME = 'reverb_soundscape_20200501' # REG_REVERB_NAME = 'reverb_soundscape_20190326_22050' # REG_REVERB_WAV_PATH = 'tests/data/regression/reverb_soundscape_20170928.wav' # REG_REVERB_JAM_PATH = 'tests/data/regression/reverb_soundscape_20170928.jams' # REG_REVERB_TXT_PATH = 'tests/data/regression/reverb_soundscape_20170928.txt' # fg and bg labels for testing FB_LABELS = ['car_horn', 'human_voice', 'siren'] BG_LABELS = ['park', 'restaurant', 'street'] SAMPLE_RATES = [22050, 44100] def test_names(name, rate, exts=('wav', 'jams', 'txt')): return [os.path.join('tests/data/regression', '{}_{}.{}'.format(name, rate, ext)) for ext in exts] for rate in SAMPLE_RATES: test_names(REG_NAME, rate) print("==========USING BELOW FOR TESTS==============") VAR_NAMES_PARTIAL = ('REG', 'REG_BGONLY', 'REG_REVERB') FILE_BASENAMES = (REG_NAME, REG_BGONLY_NAME, REG_REVERB_NAME) FILE_TYPES = ('WAV', 'JAM', 'TXT') for var, name in zip(VAR_NAMES_PARTIAL, FILE_BASENAMES): for type, path in zip(FILE_TYPES, test_names(name, rate)): print("{}_{}_PATH = '{}'".format(var, type, path)) print() print("==========USING ABOVE FOR TESTS==============") sc = scaper.Scaper(10.0, fg_path=FG_PATH, bg_path=BG_PATH) sc.ref_db = -50 sc.sr = rate # background sc.add_background( label=('const', 'park'), source_file=( 'const', 'tests/data/audio/background/park/' '268903__yonts__city-park-tel-aviv-israel.wav'), source_time=('const', 0)) # foreground events sc.add_event( label=('const', 'siren'), source_file=('const', 'tests/data/audio/foreground/' 'siren/69-Siren-1.wav'), source_time=('const', 5), event_time=('const', 2), event_duration=('const', 5), snr=('const', 5), pitch_shift=None, time_stretch=None) sc.add_event( label=('const', 'car_horn'), source_file=('const', 'tests/data/audio/foreground/' 'car_horn/17-CAR-Rolls-Royce-Horn.wav'), source_time=('const', 0), event_time=('const', 5), event_duration=('const', 2), snr=('const', 20), pitch_shift=('const', 1), time_stretch=None) sc.add_event( label=('const', 'human_voice'), source_file=('const', 'tests/data/audio/foreground/' 'human_voice/42-Human-Vocal-Voice-taxi-2_edit.wav'), source_time=('const', 0), event_time=('const', 7), event_duration=('const', 2), snr=('const', 10), pitch_shift=None, time_stretch=('const', 1.2)) wav_file, jam_file, txt_file = test_names(REG_NAME, rate) sc.generate(wav_file, jam_file, txt_path=txt_file, disable_instantiation_warnings=True) print('Wrote:', wav_file, jam_file, txt_file) wav_file, jam_file, txt_file = test_names(REG_REVERB_NAME, rate) sc.generate(wav_file, jam_file, txt_path=txt_file, reverb=0.2, disable_instantiation_warnings=True) print('Wrote:', wav_file, jam_file, txt_file) jams.load(jam_file) # soundscape with only one event will use transformer (regression test) sc = scaper.Scaper(10.0, fg_path=FG_PATH, bg_path=BG_PATH) sc.ref_db = -20 sc.sr = rate # background sc.add_background( label=('const', 'park'), source_file=('const', 'tests/data/audio/background/park/' '268903__yonts__city-park-tel-aviv-israel.wav'), source_time=('const', 0)) wav_file, jam_file, txt_file = test_names(REG_BGONLY_NAME, rate) sc.generate(wav_file, jam_file, txt_path=txt_file, reverb=0.2, disable_instantiation_warnings=True) print('Wrote:', wav_file, jam_file, txt_file)
4,773
34.626866
103
py
scaper
scaper-master/tests/__init__.py
0
0
0
py
scaper
scaper-master/tests/test_audio.py
# CREATED: 5/5/17 14:36 by Justin Salamon <justin.salamon@nyu.edu> from scaper.audio import get_integrated_lufs, match_sample_length from scaper.audio import peak_normalize from scaper.util import _close_temp_files import numpy as np import scipy.signal as sg import os import pytest from scaper.scaper_exceptions import ScaperError from pkg_resources import resource_filename import shutil import soundfile as sf import tempfile import random # fixtures SIREN_FILE = 'tests/data/audio/foreground/siren/69-Siren-1.wav' CARHORN_FILE = ( 'tests/data/audio/foreground/car_horn/17-CAR-Rolls-Royce-Horn.wav') HUMANVOICE_FILE = ( 'tests/data/audio/foreground/human_voice/' '42-Human-Vocal-Voice-all-aboard_edit.wav') DOGBARK_FILE = 'tests/data/lufs/dogbark.wav' SIREN_LUFS_I = -23.071089944980127 CARHORN_LUFS_I = -13.66146520099299 HUMANVOICE_LUFS_I = -20.061513106500225 DOGBARK_LUFS_I = -11.1952428800271 # for x4 concatenated file def test_get_integrated_lufs(): # test correct functionality audiofiles = [SIREN_FILE, CARHORN_FILE, HUMANVOICE_FILE, DOGBARK_FILE] lufsi = [SIREN_LUFS_I, CARHORN_LUFS_I, HUMANVOICE_LUFS_I, DOGBARK_LUFS_I] for af, li in zip(audiofiles, lufsi): audio, sr = sf.read(af) i = get_integrated_lufs(audio, sr) assert np.allclose(i, li) def change_format_and_subtype(audio_path): audio, sr = sf.read(audio_path) audio_info = sf.info(audio_path) formats = ['WAV', 'FLAC'] if audio_info.format in formats: formats.remove(audio_info.format) _format = random.choice(formats) subtypes = sf.available_subtypes(_format) accepted_subtypes = ['PCM_16', 'PCM_32', 'PCM_24', 'FLOAT', 'DOUBLE'] subtypes = [s for s in subtypes.keys() if s in accepted_subtypes] if audio_info.subtype in subtypes: subtypes.remove(audio_info.subtype) _subtype = random.choice(subtypes) sf.write(audio_path, audio, sr, subtype=_subtype, format=_format) def test_match_sample_length(): durations_to_match = [1, 2, 5, 7, 22500, 44100, 88200, 100001] invalid_durations_to_match = [0, -1, .5, 1.0] tmpfiles = [] with _close_temp_files(tmpfiles): carhorn = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) tmpfiles.append(carhorn) siren = tempfile.NamedTemporaryFile(suffix='.wav', delete=True) tmpfiles.append(siren) for _duration in durations_to_match: shutil.copyfile(SIREN_FILE, siren.name) shutil.copyfile(CARHORN_FILE, carhorn.name) change_format_and_subtype(siren.name) change_format_and_subtype(carhorn.name) prev_audio_info = sf.info(carhorn.name) match_sample_length(carhorn.name, _duration) carhorn_audio, _ = sf.read(carhorn.name) next_audio_info = sf.info(carhorn.name) assert carhorn_audio.shape[0] == _duration assert prev_audio_info.format == next_audio_info.format assert prev_audio_info.subtype == next_audio_info.subtype prev_audio_info = sf.info(siren.name) match_sample_length(siren.name, _duration) siren_audio, _ = sf.read(siren.name) next_audio_info = sf.info(siren.name) assert siren_audio.shape[0] == _duration assert prev_audio_info.format == next_audio_info.format assert prev_audio_info.subtype == next_audio_info.subtype # should be summable summed_events = sum([carhorn_audio, siren_audio]) assert summed_events.shape[0] == _duration for _duration in invalid_durations_to_match: pytest.raises(ScaperError, match_sample_length, carhorn.name, _duration) pytest.raises(ScaperError, match_sample_length, siren.name, _duration) def test_peak_normalize(): def sine(x, f, sr, A): return A * np.sin(2 * np.pi * f * x / sr) def square(x,f, sr, A): return A * sg.square(2 * np.pi * f * x / sr) def saw(x, f, sr, A): return A * sg.sawtooth(2 * np.pi * f * x / sr) samplerates = [16000, 44100] frequencies = [50, 100, 500, 1000, 5000, 10000, 15000, 20000] amplitudes = [0.1, 0.5, 1.0, 1.5, 2.0] event_factors = [0.5, 0.8] eps = 1e-10 # test with toy data for waveform in [sine, square, saw]: for sr in samplerates: for f in frequencies: for A in amplitudes: n_samples = sr x = np.arange(n_samples) audio = waveform(x, f, sr, A) max_sample = np.max(np.abs(audio)) estimated_scale_factor = 1.0 / (A + eps) print("\nsr: {}, f: {}, A: {}".format(sr, f, A)) print("max sample audio: {}".format(max_sample)) print("estimated scale_factor: {}".format(estimated_scale_factor)) event_audio_list = [] for factor in event_factors: event_audio_list.append(waveform(x, f, sr, A * factor)) audio_norm, event_audio_list_norm, scale_factor = \ peak_normalize(audio, event_audio_list) print("allclose on factors: {}".format(np.allclose( scale_factor, estimated_scale_factor))) print("actual scale factor: {}".format(scale_factor)) # test scale factor max_sample_audio = np.max(np.abs(audio_norm)) print("max sample audio norm: {}".format(max_sample_audio)) assert np.allclose(scale_factor, estimated_scale_factor, atol=1e-03) # test soundscape audio assert max_sample_audio < 1.0 assert np.allclose(max_sample_audio, 1.0) # test event audio for event_audio, factor in zip(event_audio_list_norm, event_factors): max_sample_event = np.max(np.abs(event_audio)) if not np.allclose( max_sample_event, A * factor * scale_factor): print(max_sample_audio, max_sample_event, A * factor * scale_factor) assert np.allclose(max_sample_event, A * factor * scale_factor, atol=1e-3)
6,611
37.219653
86
py
scaper
scaper-master/docs/conf.py
#!/usr/bin/env python3 # -*- coding: utf-8 -*- # # scaper documentation build configuration file, created by # sphinx-quickstart on Thu May 4 17:32:22 2017. # # This file is execfile()d with the current directory set to its # containing dir. # # Note that not all possible configuration values are present in this # autogenerated file. # # All configuration values have a default; values that are commented out # serve to show the default. # If extensions (or modules to document with autodoc) are in another directory, # add these directories to sys.path here. If the directory is relative to the # documentation root, use os.path.abspath to make it absolute, like shown here. import os import sys import six import sphinx_rtd_theme from sphinx.ext.autodoc import between sys.path.insert(0, os.path.abspath('.')) sys.path.insert(0, os.path.abspath('..')) # -- General configuration ------------------------------------------------ # If your documentation needs a minimal Sphinx version, state it here. # # needs_sphinx = '1.0' # Add any Sphinx extension module names here, as strings. They can be # extensions coming with Sphinx (named 'sphinx.ext.*') or your custom # ones. # extensions = [] extensions = [ 'sphinx.ext.autodoc', 'sphinx.ext.autosummary', 'sphinx.ext.coverage', 'sphinx.ext.viewcode', 'sphinx.ext.napoleon', 'sphinx_issues' ] # Link Github repo for referencing issues/PRs issues_github_path = 'justinsalamon/scaper' # Disable warnings about nonexisting document numpydoc_show_class_members = False # Add any paths that contain templates here, relative to this directory. templates_path = ['_templates'] # The suffix(es) of source filenames. # You can specify multiple suffix as a list of string: # # source_suffix = ['.rst', '.md'] source_suffix = '.rst' # The encoding of source files. # # source_encoding = 'utf-8-sig' # The master toctree document. master_doc = 'index' # General information about the project. project = 'scaper' copyright = '2017, Justin Salamon & Duncan MacConnell' author = 'Justin Salamon & Duncan MacConnell' # Mock the dependencies if six.PY3: from unittest.mock import MagicMock else: from mock import Mock as MagicMock class Mock(MagicMock): @classmethod def __getattr__(cls, name): return MagicMock() MOCK_MODULES = [ 'sox', 'jams', 'scipy', 'numpy', 'pandas', 'soundfile', ] sys.modules.update((mod_name, Mock()) for mod_name in MOCK_MODULES) # The version info for the project you're documenting, acts as replacement for # |version| and |release|, also used in various other places throughout the # built documents. # # The short X.Y version. import imp scaper_version = imp.load_source('scaper.version', '../scaper/version.py') version = scaper_version.short_version # The full version, including alpha/beta/rc tags. release = scaper_version.version # The language for content autogenerated by Sphinx. Refer to documentation # for a list of supported languages. # # This is also used if you do content translation via gettext catalogs. # Usually you set "language" from the command line for these cases. language = None # There are two options for replacing |today|: either, you set today to some # non-false value, then it is used: # # today = '' # # Else, today_fmt is used as the format for a strftime call. # # today_fmt = '%B %d, %Y' # List of patterns, relative to source directory, that match files and # directories to ignore when looking for source files. # This patterns also effect to html_static_path and html_extra_path exclude_patterns = ['_build', 'Thumbs.db', '.DS_Store'] # The reST default role (used for this markup: `text`) to use for all # documents. # # default_role = None # If true, '()' will be appended to :func: etc. cross-reference text. # # add_function_parentheses = True # If true, the current module name will be prepended to all description # unit titles (such as .. function::). # # add_module_names = True # If true, sectionauthor and moduleauthor directives will be shown in the # output. They are ignored by default. # # show_authors = False # The name of the Pygments (syntax highlighting) style to use. pygments_style = 'sphinx' # A list of ignored prefixes for module index sorting. # modindex_common_prefix = [] # If true, keep warnings as "system message" paragraphs in the built documents. # keep_warnings = False # If true, `todo` and `todoList` produce output, else they produce nothing. todo_include_todos = False # -- Options for HTML output ---------------------------------------------- # The theme to use for HTML and HTML Help pages. See the documentation for # a list of builtin themes. # # html_theme = 'alabaster' # html_theme = 'classic' html_theme = "sphinx_rtd_theme" html_theme_path = [sphinx_rtd_theme.get_html_theme_path()] # Theme options are theme-specific and customize the look and feel of a theme # further. For a list of options available for each theme, see the # documentation. # # html_theme_options = {} # Add any paths that contain custom themes here, relative to this directory. # html_theme_path = [] # The name for this set of Sphinx documents. # "<project> v<release> documentation" by default. # # html_title = 'scaper v0.1' # A shorter title for the navigation bar. Default is the same as html_title. # # html_short_title = None # The name of an image file (relative to this directory) to place at the top # of the sidebar. # # html_logo = None # The name of an image file (relative to this directory) to use as a favicon of # the docs. This file should be a Windows icon file (.ico) being 16x16 or 32x32 # pixels large. # # html_favicon = None # Add any paths that contain custom static files (such as style sheets) here, # relative to this directory. They are copied after the builtin static files, # so a file named "default.css" will overwrite the builtin "default.css". html_static_path = ['_static'] # Add any extra paths that contain custom files (such as robots.txt or # .htaccess) here, relative to this directory. These files are copied # directly to the root of the documentation. # # html_extra_path = [] # If not None, a 'Last updated on:' timestamp is inserted at every page # bottom, using the given strftime format. # The empty string is equivalent to '%b %d, %Y'. # # html_last_updated_fmt = None # If true, SmartyPants will be used to convert quotes and dashes to # typographically correct entities. # # html_use_smartypants = True # Custom sidebar templates, maps document names to template names. # # html_sidebars = {} # Additional templates that should be rendered to pages, maps page names to # template names. # # html_additional_pages = {} # If false, no module index is generated. # # html_domain_indices = True # If false, no index is generated. # # html_use_index = True # If true, the index is split into individual pages for each letter. # # html_split_index = False # If true, links to the reST sources are added to the pages. # # html_show_sourcelink = True # If true, "Created using Sphinx" is shown in the HTML footer. Default is True. # # html_show_sphinx = True # If true, "(C) Copyright ..." is shown in the HTML footer. Default is True. # # html_show_copyright = True # If true, an OpenSearch description file will be output, and all pages will # contain a <link> tag referring to it. The value of this option must be the # base URL from which the finished HTML is served. # # html_use_opensearch = '' # This is the file name suffix for HTML files (e.g. ".xhtml"). # html_file_suffix = None # Language to be used for generating the HTML full-text search index. # Sphinx supports the following languages: # 'da', 'de', 'en', 'es', 'fi', 'fr', 'h', 'it', 'ja' # 'nl', 'no', 'pt', 'ro', 'r', 'sv', 'tr', 'zh' # # html_search_language = 'en' # A dictionary with options for the search language support, empty by default. # 'ja' uses this config value. # 'zh' user can custom change `jieba` dictionary path. # # html_search_options = {'type': 'default'} # The name of a javascript file (relative to the configuration directory) that # implements a search results scorer. If empty, the default will be used. # # html_search_scorer = 'scorer.js' # Output file base name for HTML help builder. htmlhelp_basename = 'scaperdoc' # -- Options for LaTeX output --------------------------------------------- latex_elements = { # The paper size ('letterpaper' or 'a4paper'). # # 'papersize': 'letterpaper', # The font size ('10pt', '11pt' or '12pt'). # # 'pointsize': '10pt', # Additional stuff for the LaTeX preamble. # # 'preamble': '', # Latex figure (float) alignment # # 'figure_align': 'htbp', } # Grouping the document tree into LaTeX files. List of tuples # (source start file, target name, title, # author, documentclass [howto, manual, or own class]). latex_documents = [ (master_doc, 'scaper.tex', 'scaper Documentation', 'Justin Salamon \\& Duncan MacConnell', 'manual'), ] # The name of an image file (relative to this directory) to place at the top of # the title page. # # latex_logo = None # For "manual" documents, if this is true, then toplevel headings are parts, # not chapters. # # latex_use_parts = False # If true, show page references after internal links. # # latex_show_pagerefs = False # If true, show URL addresses after external links. # # latex_show_urls = False # Documents to append as an appendix to all manuals. # # latex_appendices = [] # It false, will not define \strong, \code, itleref, \crossref ... but only # \sphinxstrong, ..., \sphinxtitleref, ... To help avoid clash with user added # packages. # # latex_keep_old_macro_names = True # If false, no module index is generated. # # latex_domain_indices = True # -- Options for manual page output --------------------------------------- # One entry per manual page. List of tuples # (source start file, name, description, authors, manual section). man_pages = [ (master_doc, 'scaper', 'scaper Documentation', [author], 1) ] # If true, show URL addresses after external links. # # man_show_urls = False # -- Options for Texinfo output ------------------------------------------- # Grouping the document tree into Texinfo files. List of tuples # (source start file, target name, title, author, # dir menu entry, description, category) texinfo_documents = [ (master_doc, 'scaper', 'scaper Documentation', author, 'scaper', 'One line description of project.', 'Miscellaneous'), ] # Documents to append as an appendix to all manuals. # # texinfo_appendices = [] # If false, no module index is generated. # # texinfo_domain_indices = True # How to display URL addresses: 'footnote', 'no', or 'inline'. # # texinfo_show_urls = 'footnote' # If true, do not generate a @detailmenu in the "Top" node's menu. # # texinfo_no_detailmenu = False def maybe_skip_member(app, what, name, obj, skip, options): # print(app) # print(what) # print(name) # print(obj) # print(skip) # print(options) if name == 'EventSpec' or name[0] == '_': return True else: return False def setup(app): app.connect('autodoc-skip-member', maybe_skip_member)
11,254
27.20802
80
py
scaper
scaper-master/scaper/scaper_warnings.py
#!/usr/bin/env python # -*- encoding: utf-8 -*- # CREATED: 10/13/16 7:08 PM by Justin Salamon <justin.salamon@nyu.edu> '''Warning classes for Scaper''' class ScaperWarning(Warning): '''The root Scaper warning class''' pass
235
18.666667
70
py
scaper
scaper-master/scaper/core.py
try: import soxbindings as sox except: # pragma: no cover import sox # pragma: no cover import soundfile import os import warnings import jams from collections import namedtuple import logging import tempfile import numpy as np import shutil import csv from copy import deepcopy from .scaper_exceptions import ScaperError from .scaper_warnings import ScaperWarning from .util import _close_temp_files from .util import _set_temp_logging_level from .util import _get_sorted_files from .util import _validate_folder_path from .util import _populate_label_list from .util import _check_random_state from .util import _sample_trunc_norm from .util import _sample_uniform from .util import _sample_choose from .util import _sample_choose_weighted from .util import _sample_normal from .util import _sample_const from .util import max_polyphony from .util import polyphony_gini from .util import is_real_number, is_real_array from .audio import get_integrated_lufs from .audio import peak_normalize from .version import version as scaper_version # HEADS UP! Adding a new distribution tuple? # Make sure it's properly handled in all of the following: # SUPPORTED_DIST # _validate_distribution # _validate_time # _validate_duration # _validate_time_stretch # _validate_pitch_shift # _validate_label # _validate_source_file # _validate_snr # _ensure_satisfiable_source_time_tuple # _instantiate_event # add_background (docstring) # add_event (docstring) SUPPORTED_DIST = {"const": _sample_const, "choose": _sample_choose, "choose_weighted": _sample_choose_weighted, "uniform": _sample_uniform, "normal": _sample_normal, "truncnorm": _sample_trunc_norm} # Define single event spec as namedtuple EventSpec = namedtuple( 'EventSpec', ['label', 'source_file', 'source_time', 'event_time', 'event_duration', 'snr', 'role', 'pitch_shift', 'time_stretch']) ''' Container for storing event specifications, either probabilistic (i.e. using distribution tuples to specify possible values) or instantiated (i.e. storing constants directly). ''' def generate_from_jams(jams_infile, audio_outfile=None, fg_path=None, bg_path=None, jams_outfile=None, save_isolated_events=False, isolated_events_path=None, disable_sox_warnings=True, txt_path=None, txt_sep='\t'): ''' Generate a soundscape based on an existing scaper JAMS file and return as an audio file, a JAMS annotation, a simplified annotation list, and a list containing the audio samples of each individual background and foreground event. If output paths are provided, these objects will also be saved to disk. Parameters ---------- jams_infile : str Path to JAMS file (must be a file previously generated by Scaper). audio_outfile : str Path for saving the generated soundscape audio. fg_path : str or None Specifies a different path for foreground audio than the one stored in the input jams file. For the reconstruction to be successful the folder and file structure inside this path must be identical the one that was used to create the input jams file. If None (default), the fg_path from the input jams file will be used. bg_path : str or None Specifies a different path for background audio than the one stored in the input jams file. For the reconstruction to be successful the folder and file structure inside this path must be identical the one that was used to create the input jams file. If None (default), the bg_path from the input jams file will be used. jams_outfile : str or None Path for saving new JAMS file, if None (default) a new JAMS is not saved. Useful when either fg_path or bg_path is not None, as it saves a new JAMS files where the source file paths match the new fg_path and/or bg_path. save_isolated_events : bool If True, this will save the isolated event audio in a directory adjacent to the generated soundscape mixture, or to the path defined by `isolated_events_path`. The audio of the isolated events sum up to the mixture if reverb is not applied. Isolated events can be found (by default) at `<audio_outfile parent folder>/<audio_outfile name>_events`. Isolated event file names follow the pattern: `<role><idx>_<label>`, where idx is the index of the isolated event in self.fg_spec or self.bg_spec (this allows events of the same label to be added more than once to the soundscape without breaking things). Role is "background" or "foreground". For example: `foreground0_siren.wav` or `background0_park.wav`. isolated_events_path : str Path to folder for saving isolated events. If None, defaults to `<audio_outfile parent folder>/<audio_outfile name>_events`. disable_sox_warnings : bool When True (default), warnings from the pysox module are suppressed unless their level is ``'CRITICAL'``. If you're experiencing issues related to audio I/O setting this parameter to False may help with debugging. txt_path: str or None Path for saving a simplified annotation in a space separated format [onset offset label] where onset and offset are in seconds. Good for loading labels in e.g. Audacity. If None, does not save txt annotation to disk. txt_sep: str The separator to use when saving a simplified annotation as a text file (default is tab for compatibility with Audacity label files). Only relevant if txt_path is not None. Returns ------- soundscape_audio : np.ndarray The audio samples of the generated soundscape. Returns None if no_audio=True. soundscape_jam: jams.JAMS The JAMS object containing the full soundscape annotation. annotation_list : list A simplified annotation in a space-separated format [onset offset label] where onset and offset are in seconds. event_audio_list: list A list of np.ndarrays containing the audio samples of every individual background and foreground sound event. Events are listed in the same order in which they appear in the jams annotations data list, and can be matched with: `for obs, event_audio in zip(ann.data, event_audio_list): ...`. Raises ------ ScaperError If jams_infile does not point to a valid JAMS file that was previously generated by Scaper and contains an annotation of the scaper namespace. ''' soundscape_jam = jams.load(jams_infile) anns = soundscape_jam.search(namespace='scaper') if len(anns) == 0: raise ScaperError( 'JAMS file does not contain any annotation with namespace ' 'scaper.') ann = soundscape_jam.annotations.search(namespace='scaper')[0] # Update paths if fg_path is None: new_fg_path = ann.sandbox.scaper['fg_path'] else: new_fg_path = os.path.expanduser(fg_path) # Update source files for obs in ann.data: if obs.value['role'] == 'foreground': sourcefile = obs.value['source_file'] sourcefilename = os.path.basename(sourcefile) parent = os.path.dirname(sourcefile) parentname = os.path.basename(parent) newsourcefile = os.path.join( new_fg_path, parentname, sourcefilename) obs.value['source_file'] = newsourcefile # hacky # Update sandbox ann.sandbox.scaper['fg_path'] = new_fg_path if bg_path is None: new_bg_path = ann.sandbox.scaper['bg_path'] else: new_bg_path = os.path.expanduser(bg_path) # Update source files for obs in ann.data: if obs.value['role'] == 'background': sourcefile = obs.value['source_file'] sourcefilename = os.path.basename(sourcefile) parent = os.path.dirname(sourcefile) parentname = os.path.basename(parent) newsourcefile = os.path.join( new_bg_path, parentname, sourcefilename) obs.value['source_file'] = newsourcefile # hacky # Update sandbox ann.sandbox.scaper['bg_path'] = new_bg_path # Create scaper object if 'original_duration' in ann.sandbox.scaper: duration = ann.sandbox.scaper['original_duration'] else: duration = ann.sandbox.scaper['duration'] warnings.warn( "Couldn't find original_duration field in the scaper sandbox, " "using duration field instead. This can lead to incorrect behavior " "if generating from a jams file that has been trimmed previously.", ScaperWarning) protected_labels = ann.sandbox.scaper['protected_labels'] sc = Scaper(duration, new_fg_path, new_bg_path, protected_labels) # Set synthesis parameters if 'sr' in ann.sandbox.scaper: # backwards compatibility sc.sr = ann.sandbox.scaper['sr'] sc.ref_db = ann.sandbox.scaper['ref_db'] sc.n_channels = ann.sandbox.scaper['n_channels'] sc.fade_in_len = ann.sandbox.scaper['fade_in_len'] sc.fade_out_len = ann.sandbox.scaper['fade_out_len'] # Pull generation parameters from annotation reverb = ann.sandbox.scaper['reverb'] if 'fix_clipping' in ann.sandbox.scaper.keys(): fix_clipping = ann.sandbox.scaper['fix_clipping'] else: fix_clipping = False if 'peak_normalization' in ann.sandbox.scaper.keys(): peak_normalization = ann.sandbox.scaper['peak_normalization'] else: peak_normalization = False if 'quick_pitch_time' in ann.sandbox.scaper.keys(): quick_pitch_time = ann.sandbox.scaper['quick_pitch_time'] else: quick_pitch_time = False # Cast ann.sandbox.scaper to a Sandbox object ann.sandbox.scaper = jams.Sandbox(**ann.sandbox.scaper) # Generate audio soundscape_audio, event_audio_list, scale_factor, ref_db_change = \ sc._generate_audio(audio_outfile, ann, reverb=reverb, fix_clipping=fix_clipping, peak_normalization=peak_normalization, quick_pitch_time=quick_pitch_time, save_isolated_events=save_isolated_events, isolated_events_path=isolated_events_path, disable_sox_warnings=disable_sox_warnings) # TODO: Stick to heavy handed overwriting for now, in the future we # should consolidate this with what happens inside _instantiate(). ann.sandbox.scaper.reverb = reverb ann.sandbox.scaper.fix_clipping = fix_clipping ann.sandbox.scaper.peak_normalization = peak_normalization ann.sandbox.scaper.quick_pitch_time = quick_pitch_time ann.sandbox.scaper.save_isolated_events = save_isolated_events ann.sandbox.scaper.isolated_events_path = isolated_events_path ann.sandbox.scaper.disable_sox_warnings = disable_sox_warnings ann.sandbox.scaper.peak_normalization_scale_factor = scale_factor ann.sandbox.scaper.ref_db_change = ref_db_change ann.sandbox.scaper.ref_db_generated = sc.ref_db + ref_db_change # If there are slice (trim) operations, need to perform them! # Need to add this logic for the isolated events too. if 'slice' in ann.sandbox.keys(): for sliceop in ann.sandbox['slice']: # must use temp file in order to save to same file tmpfiles = [] audio_files = [audio_outfile] + ann.sandbox.scaper.isolated_events_audio_path with _close_temp_files(tmpfiles): for audio_file in audio_files: # Create tmp file tmpfiles.append( tempfile.NamedTemporaryFile(suffix='.wav', delete=False)) # Save trimmed result to temp file tfm = sox.Transformer() tfm.trim(sliceop['slice_start'], sliceop['slice_end']) tfm.build(audio_file, tmpfiles[-1].name) # Copy result back to original file shutil.copyfile(tmpfiles[-1].name, audio_file) # Optionally save new jams file if jams_outfile is not None: soundscape_jam.save(jams_outfile) # Create annotation list annotation_list = [] for obs in ann.data: if obs.value['role'] == 'foreground': annotation_list.append( [obs.time, obs.time + obs.duration, obs.value['label']]) if txt_path is not None: with open(txt_path, 'w') as csv_file: writer = csv.writer(csv_file, delimiter=txt_sep) writer.writerows(annotation_list) return soundscape_audio, soundscape_jam, annotation_list, event_audio_list def trim(audio_infile, jams_infile, audio_outfile, jams_outfile, start_time, end_time, no_audio=False): ''' Trim an audio file and corresponding Scaper JAMS file and save to disk. Given an input audio file and corresponding jams file, trim both the audio and all annotations in the jams file to the time range ``[start_time, end_time]`` and save the result to ``audio_outfile`` and ``jams_outfile`` respectively. This function uses ``jams.slice()`` for trimming the jams file while ensuring the start times of the jam's annotations and observations they contain match the trimmed audio file. Parameters ---------- audio_infile : str Path to input audio file jams_infile : str Path to input jams file audio_outfile : str Path to output trimmed audio file jams_outfile : str Path to output trimmed jams file start_time : float Start time for trimmed audio/jams end_time : float End time for trimmed audio/jams no_audio : bool If true, operates on the jams only. Audio input and output paths don't have to point to valid files. ''' # First trim jams (might raise an error) jam = jams.load(jams_infile) jam_sliced = jam.slice(start_time, end_time, strict=False) # Special work for annotations of the scaper 'scaper' namespace for ann in jam_sliced.annotations: if ann.namespace == 'scaper': # DON'T MODIFY event's value dict! Keeps original instantiated # values for reconstruction / reproducibility. # Count number of FG events n_events = 0 for obs in ann.data: if obs.value['role'] == 'foreground': n_events += 1 # Re-compute max polyphony poly = max_polyphony(ann) # Re-compute polyphony gini gini = polyphony_gini(ann) # Update specs in sandbox ann.sandbox.scaper['n_events'] = n_events ann.sandbox.scaper['polyphony_max'] = poly ann.sandbox.scaper['polyphony_gini'] = gini ann.sandbox.scaper['duration'] = ann.duration # Save result to output jams file jam_sliced.save(jams_outfile) # Next, trim audio if not no_audio: tfm = sox.Transformer() tfm.trim(start_time, end_time) if audio_outfile != audio_infile: tfm.build(audio_infile, audio_outfile) else: # must use temp file in order to save to same file tmpfiles = [] with _close_temp_files(tmpfiles): # Create tmp file tmpfiles.append( tempfile.NamedTemporaryFile( suffix='.wav', delete=False)) # Save trimmed result to temp file tfm.build(audio_infile, tmpfiles[-1].name) # Copy result back to original file shutil.copyfile(tmpfiles[-1].name, audio_outfile) def _get_value_from_dist(dist_tuple, random_state): ''' Sample a value from the provided distribution tuple. Given a distribution tuple, validate its format/values and then sample and return a single value from the distribution specified by the tuple. Parameters ---------- dist_tuple : tuple Distribution tuple to be validated. See ``Scaper.add_event`` for details about the expected format for the distribution tuple. Returns ------- value A value from the specified distribution. See Also -------- Scaper.add_event : Add a foreground sound event to the foreground specification. _validate_distribution : Check whether a tuple specifying a parameter distribution has a valid format, if not raise an error. ''' # Make sure it's a valid distribution tuple _validate_distribution(dist_tuple) return SUPPORTED_DIST[dist_tuple[0]](*dist_tuple[1:], random_state=random_state) def _validate_distribution(dist_tuple): ''' Check whether a tuple specifying a parameter distribution has a valid format, if not raise an error. Parameters ---------- dist_tuple : tuple Tuple specifying a distribution to sample from. See Scaper.add_event for details about the expected format of the tuple and allowed values. Raises ------ ScaperError If the tuple does not have a valid format. See Also -------- Scaper.add_event : Add a foreground sound event to the foreground specification. ''' # Make sure it's a tuple if not isinstance(dist_tuple, tuple): raise ScaperError('Distribution tuple must be of type tuple.') # Make sure the tuple contains at least 2 items if len(dist_tuple) < 2: raise ScaperError('Distribution tuple must be at least of length 2.') # Make sure the first item is one of the supported distribution names if dist_tuple[0] not in SUPPORTED_DIST.keys(): raise ScaperError( "Unsupported distribution name: {:s}".format(dist_tuple[0])) # If it's a constant distribution, tuple must be of length 2 if dist_tuple[0] == 'const': if len(dist_tuple) != 2: raise ScaperError('"const" distribution tuple must be of length 2') # If it's a choose, tuple must be of length 2 and second item of type list elif dist_tuple[0] == 'choose': if len(dist_tuple) != 2 or not isinstance(dist_tuple[1], list): raise ScaperError( 'The "choose" distribution tuple must be of length 2 where ' 'the second item is a list.') # If it's a choose_weighted, tuple must be of length 3, items 2 and 3 must # be lists of the same length, and the list in item 3 must contain floats # in the range [0, 1] that sum to 1 (i.e. valid probabilities). elif dist_tuple[0] == 'choose_weighted': if len(dist_tuple) != 3: raise ScaperError('"choose_weighted" distribution tuple must have length 3') if not isinstance(dist_tuple[1], list) or \ not isinstance(dist_tuple[2], list) or \ len(dist_tuple[1]) != len(dist_tuple[2]): msg = ('The 2nd and 3rd items of the "choose_weighted" distribution tuple ' 'must be lists of the same length.') raise ScaperError(msg) probabilities = np.asarray(dist_tuple[2]) if probabilities.min() < 0 or probabilities.max() > 1: msg = ('Values in the probabilities list of the "choose_weighted" ' 'distribution tuple must be in the range [0, 1].') raise ScaperError(msg) if not np.allclose(probabilities.sum(), 1): msg = ('Values in the probabilities list of the "choose_weighted" ' 'distribution tuple must sum to 1.') raise ScaperError(msg) # If it's a uniform distribution, tuple must be of length 3, 2nd item must # be a real number and 3rd item must be real and greater/equal to the 2nd. elif dist_tuple[0] == 'uniform': if (len(dist_tuple) != 3 or not is_real_number(dist_tuple[1]) or not is_real_number(dist_tuple[2]) or dist_tuple[1] > dist_tuple[2]): raise ScaperError( 'The "uniform" distribution tuple be of length 2, where the ' '2nd item is a real number and the 3rd item is a real number ' 'and greater/equal to the 2nd item.') # If it's a normal distribution, tuple must be of length 3, 2nd item must # be a real number and 3rd item must be a non-negative real elif dist_tuple[0] == 'normal': if (len(dist_tuple) != 3 or not is_real_number(dist_tuple[1]) or not is_real_number(dist_tuple[2]) or dist_tuple[2] < 0): raise ScaperError( 'The "normal" distribution tuple must be of length 3, where ' 'the 2nd item (mean) is a real number and the 3rd item (std ' 'dev) is real and non-negative.') elif dist_tuple[0] == 'truncnorm': if (len(dist_tuple) != 5 or not is_real_number(dist_tuple[1]) or not is_real_number(dist_tuple[2]) or not is_real_number(dist_tuple[3]) or not is_real_number(dist_tuple[4]) or dist_tuple[2] < 0 or dist_tuple[4] < dist_tuple[3]): raise ScaperError( 'The "truncnorm" distribution tuple must be of length 5, ' 'where the 2nd item (mean) is a real number, the 3rd item ' '(std dev) is real and non-negative, the 4th item (trunc_min) ' 'is a real number and the 5th item (trun_max) is a real ' 'number that is equal to or greater than trunc_min.') def _ensure_satisfiable_source_time_tuple(source_time, source_duration, event_duration): ''' Modify a source_time distribution tuple according to the duration of the source and the duration of the event. This allows you to sample from anywhere in a source file without knowing the exact duration of every source file. Parameters ---------- source_time : tuple Tuple specifying a distribution to sample from. See Scaper.add_event for details about the expected format of the tuple and allowed values. source_duration : float Duration of the source audio file. event_duration : float Duration of the event to be extracted from the source file. See Also -------- Scaper.add_event : Add a foreground sound event to the foreground specification. ''' _validate_distribution(source_time) old_source_time = deepcopy(source_time) source_time = list(source_time) # If it's a constant distribution, just make sure it's within bounds. if source_time[0] == 'const': if source_time[1] + event_duration > source_duration: source_time[1] = max(0, source_duration - event_duration) # If it's a choose, iterate through the list to make sure it's all in bounds. # Some logic here so we don't add stuff out of bounds more than once. elif source_time[0] == 'choose': for i, t in enumerate(source_time[1]): if t + event_duration > source_duration: source_time[1][i] = max(0, source_duration - event_duration) source_time[1] = list(set(source_time[1])) # For weighted_choose we do the same as choose but without removing duplicates elif source_time[0] == 'choose_weighted': for i, t in enumerate(source_time[1]): if t + event_duration > source_duration: source_time[1][i] = max(0, source_duration - event_duration) # If it's a uniform distribution, tuple must be of length 3, We change the 3rd # item to source_duration - event_duration so that we stay in bounds. If the min # out of bounds, we change it to be source_duration - event_duration. elif source_time[0] == 'uniform': if source_time[1] + event_duration > source_duration: source_time[1] = max(0, source_duration - event_duration) if source_time[2] + event_duration > source_duration: source_time[2] = max(0, source_duration - event_duration) if (source_time[1] == source_time[2]): # switch to const source_time = ['const', source_time[1]] # If it's a normal distribution, we change the mean of the distribution to # source_duration - event_duration if source_duration - mean < event_duration. elif source_time[0] == 'normal': if source_time[1] + event_duration > source_duration: source_time[1] = max(0, source_duration - event_duration) # If it's a truncated normal distribution, we change the mean as we did above for a # normal distribution, and change the max (5th item) to # source_duration - event_duration if it's bigger. If the min is out of bounds, we # change it like in the uniform case. elif source_time[0] == 'truncnorm': if source_time[1] + event_duration > source_duration: source_time[1] = max(0, source_duration - event_duration) if source_time[3] + event_duration > source_duration: source_time[3] = max(0, source_duration - event_duration) if source_time[4] + event_duration > source_duration: source_time[4] = max(0, source_duration - event_duration) if (source_time[3] == source_time[4]): # switch to const source_time = ['const', source_time[1]] source_time = tuple(source_time) # check if the source_time changed from the old_source_time to throw a warning. # it gets set here but the warning happens after the return from this call warn = (source_time != old_source_time) return tuple(source_time), warn def _validate_label(label, allowed_labels): ''' Validate that a label tuple is in the right format and that it's values are valid. Parameters ---------- label : tuple Label tuple (see ```Scaper.add_event``` for required format). allowed_labels : list List of allowed labels. Raises ------ ScaperError If the validation fails. ''' # Make sure it's a valid distribution tuple _validate_distribution(label) # Make sure it's one of the allowed distributions for a label and that the # label value is one of the allowed labels. if label[0] == "const": if not label[1] in allowed_labels: raise ScaperError( 'Label value must match one of the available labels: ' '{:s}'.format(str(allowed_labels))) elif label[0] == "choose" or label[0] == "choose_weighted": if label[1]: # list is not empty if not set(label[1]).issubset(set(allowed_labels)): raise ScaperError( 'Label list provided must be a subset of the available ' 'labels: {:s}'.format(str(allowed_labels))) else: raise ScaperError( 'Label must be specified using a "const" or "choose" tuple.') def _validate_source_file(source_file_tuple, label_tuple): ''' Validate that a source_file tuple is in the right format a that it's values are valid. Parameters ---------- source_file : tuple Source file tuple (see ```Scaper.add_event``` for required format). label : str Label tuple (see ```Scaper.add_event``` for required format). Raises ------ ScaperError If the validation fails. ''' # Make sure it's a valid distribution tuple _validate_distribution(source_file_tuple) _validate_distribution(label_tuple) # If source file is specified explicitly if source_file_tuple[0] == "const": # 1. the filepath must point to an existing file if not os.path.isfile(source_file_tuple[1]): raise ScaperError( "Source file not found: {:s}".format(source_file_tuple[1])) # 2. the label must match the file's parent folder name parent_name = os.path.basename(os.path.dirname(source_file_tuple[1])) if label_tuple[0] != "const" or label_tuple[1] != parent_name: raise ScaperError( "Source file's parent folder name does not match label.") # Otherwise it must be specified using one of "choose" or "choose_weighted" elif source_file_tuple[0] == "choose" or source_file_tuple[0] == "choose_weighted": if source_file_tuple[1]: # list is not empty if not all(os.path.isfile(x) for x in source_file_tuple[1]): raise ScaperError( 'Source file list must either be empty or all paths in ' 'the list must point to valid files.') else: raise ScaperError( 'Source file must be specified using a "const" or "choose" tuple.') def _validate_time(time_tuple): ''' Validate that a time tuple has the right format and that the specified distribution cannot result in a negative time. Parameters ---------- time_tuple : tuple Time tuple (see ```Scaper.add_event``` for required format). Raises ------ ScaperError If the validation fails. ''' # Make sure it's a valid distribution tuple _validate_distribution(time_tuple) # Ensure the values are valid for time if time_tuple[0] == "const": if (time_tuple[1] is None or not is_real_number(time_tuple[1]) or time_tuple[1] < 0): raise ScaperError( 'Time must be a real non-negative number.') elif time_tuple[0] == "choose" or time_tuple[0] == "choose_weighted": if (not time_tuple[1] or not is_real_array(time_tuple[1]) or not all(x is not None for x in time_tuple[1]) or not all(x >= 0 for x in time_tuple[1])): raise ScaperError( 'Time list must be a non-empty list of non-negative real ' 'numbers.') elif time_tuple[0] == "uniform": if time_tuple[1] < 0: raise ScaperError( 'A "uniform" distribution tuple for time must have ' 'min_value >= 0') elif time_tuple[0] == "normal": warnings.warn( 'A "normal" distribution tuple for time can result in ' 'negative values, in which case the distribution will be ' 're-sampled until a positive value is returned: this can result ' 'in an infinite loop!', ScaperWarning) elif time_tuple[0] == "truncnorm": if time_tuple[3] < 0: raise ScaperError( 'A "truncnorm" distirbution tuple for time must specify a non-' 'negative trunc_min value.') def _validate_duration(duration_tuple): ''' Validate that a duration tuple has the right format and that the specified distribution cannot result in a negative or zero value. Parameters ---------- duration : tuple Duration tuple (see ```Scaper.add_event``` for required format). Raises ------ ScaperError If the validation fails. ''' # Make sure it's a valid distribution tuple _validate_distribution(duration_tuple) # Ensure the values are valid for duration if duration_tuple[0] == "const": if (not is_real_number(duration_tuple[1]) or duration_tuple[1] <= 0): raise ScaperError( 'Duration must be a real number greater than zero.') elif duration_tuple[0] == "choose" or duration_tuple[0] == "choose_weighted": if (not duration_tuple[1] or not is_real_array(duration_tuple[1]) or not all(x > 0 for x in duration_tuple[1])): raise ScaperError( 'Duration list must be a non-empty list of positive real ' 'numbers.') elif duration_tuple[0] == "uniform": if duration_tuple[1] <= 0: raise ScaperError( 'A "uniform" distribution tuple for duration must have ' 'min_value > 0') elif duration_tuple[0] == "normal": warnings.warn( 'A "normal" distribution tuple for duration can result in ' 'non-positives values, in which case the distribution will be ' 're-sampled until a positive value is returned: this can result ' 'in an infinite loop!', ScaperWarning) elif duration_tuple[0] == "truncnorm": if duration_tuple[3] <= 0: raise ScaperError( 'A "truncnorm" distirbution tuple for time must specify a ' 'positive trunc_min value.') def _validate_snr(snr_tuple): ''' Validate that an snr distribution tuple has the right format. Parameters ---------- snr : tuple SNR tuple (see ```Scaper.add_event``` for required format). Raises ------ ScaperError If the validation fails. ''' # Make sure it's a valid distribution tuple _validate_distribution(snr_tuple) # Ensure the values are valid for SNR if snr_tuple[0] == "const": if not is_real_number(snr_tuple[1]): raise ScaperError( 'SNR must be a real number.') elif snr_tuple[0] == "choose" or snr_tuple[0] == "choose_weighted": if (not snr_tuple[1] or not is_real_array(snr_tuple[1])): raise ScaperError( 'SNR list must be a non-empty list of real numbers.') # No need to check for "uniform" and "normal" since they must produce a # real number and technically speaking any real number is a valid SNR. # TODO: do we want to impose limits on the possible SNR values? def _validate_pitch_shift(pitch_shift_tuple): ''' Validate that a pitch_shift distribution tuple has the right format. Parameters ---------- pitch_shift_tuple : tuple Pitch shift tuple (see ```Scaper.add_event``` for required format). Raises ------ ScaperError If the validation fails. ''' # If the tuple is none then it's valid if pitch_shift_tuple is not None: # Make sure it's a valid distribution tuple _validate_distribution(pitch_shift_tuple) # Ensure the values are valid for pitch shift if pitch_shift_tuple[0] == "const": if not is_real_number(pitch_shift_tuple[1]): raise ScaperError( 'Pitch shift must be a real number.') elif pitch_shift_tuple[0] == "choose" or pitch_shift_tuple[0] == "choose_weighted": if (not pitch_shift_tuple[1] or not is_real_array(pitch_shift_tuple[1])): raise ScaperError( 'Pitch shift list must be a non-empty list of real ' 'numbers.') # No need to check for "uniform" and "normal" since they must produce a # real number and technically speaking any real number is a valid pitch # shift # TODO: do we want to impose limits on the possible pitch shift values? def _validate_time_stretch(time_stretch_tuple): ''' Validate that a time_stretch distribution tuple has the right format. Parameters ---------- time_stretch_tuple: tuple Time stretch tuple (see ```Scaper.add_event``` for required format). Raises ------ ScaperError If the validation fails. ''' # if the tuple is none then its valid if time_stretch_tuple is not None: # Make sure it's a valid distribution tuple _validate_distribution(time_stretch_tuple) # Ensure the values are valid for time stretch if time_stretch_tuple[0] == "const": if (not is_real_number(time_stretch_tuple[1]) or time_stretch_tuple[1] <= 0): raise ScaperError( 'Time stretch must be a real number greater than zero.') elif time_stretch_tuple[0] == "choose" or time_stretch_tuple[0] == "choose_weighted": if (not time_stretch_tuple[1] or not is_real_array(time_stretch_tuple[1]) or not all(x > 0 for x in time_stretch_tuple[1])): raise ScaperError( 'Time stretch list must be a non-empty list of positive ' 'real numbers.') elif time_stretch_tuple[0] == "uniform": if time_stretch_tuple[1] <= 0: raise ScaperError( 'A "uniform" distribution tuple for time stretch must have ' 'min_value > 0') elif time_stretch_tuple[0] == "normal": warnings.warn( 'A "normal" distribution tuple for time stretch can result in ' 'non-positives values, in which case the distribution will be ' 're-sampled until a positive value is returned: this can ' 'result in an infinite loop!', ScaperWarning) elif time_stretch_tuple[0] == "truncnorm": if time_stretch_tuple[3] <= 0: raise ScaperError( 'A "truncnorm" distirbution tuple for time stretch must ' 'specify a positive trunc_min value.') # TODO: do we want to impose limits on the possible time stretch # values? def _validate_event(label, source_file, source_time, event_time, event_duration, snr, allowed_labels, pitch_shift, time_stretch): ''' Check that event parameter values are valid. Parameters ---------- label : tuple source_file : tuple source_time : tuple event_time : tuple event_duration : tuple snr : tuple allowed_labels : list List of allowed labels for the event. pitch_shift : tuple or None time_stretch: tuple or None Raises ------ ScaperError : If any of the input parameters has an invalid format or value. See Also -------- Scaper.add_event : Add a foreground sound event to the foreground specification. ''' # allowed_labels must be a list. All other parameters will be validated # individually. if not isinstance(allowed_labels, list): raise ScaperError('allowed_labels must be of type list.') # SOURCE FILE _validate_source_file(source_file, label) # LABEL _validate_label(label, allowed_labels) # SOURCE TIME _validate_time(source_time) # EVENT TIME _validate_time(event_time) # EVENT DURATION _validate_duration(event_duration) # SNR _validate_snr(snr) # Pitch shift _validate_pitch_shift(pitch_shift) # Time stretch _validate_time_stretch(time_stretch) class Scaper(object): ''' Create a Scaper object. Parameters ---------- duration : float Duration of the soundscape, in seconds. fg_path : str Path to foreground folder. bg_path : str Path to background folder. protected_labels : list Provide a list of protected foreground labels. When a foreground label is in the protected list it means that when a sound event matching the label gets added to a soundscape instantiation the duration of the source audio file cannot be altered, and the duration value that was provided in the specification will be ignored. Adding labels to the protected list is useful for sound events whose semantic validity would be lost if the sound were trimmed before the sound event ends, for example an animal vocalization such as a dog bark. random_state : int, RandomState instance or None, optional (default=None) If int, random_state is the seed used by the random number generator; If RandomState instance, random_state is the random number generator; If None, the random number generator is the RandomState instance used by np.random. Note that if the random state is passed as a RandomState instance, it is passed by reference, not value. This will lead to the Scaper object advancing the state of the random state object if you use it elsewhere. ''' def __init__(self, duration, fg_path, bg_path, protected_labels=[], random_state=None): ''' Create a Scaper object. Parameters ---------- duration : float Duration of the soundscape, in seconds. fg_path : str Path to foreground folder. bg_path : str Path to background folder. protected_labels : list Provide a list of protected foreground labels. When a foreground label is in the protected list it means that when a sound event matching the label gets added to a soundscape instantiation the duration of the source audio file cannot be altered, and the duration value that was provided in the specification will be ignored. Adding labels to the protected list is useful for sound events whose semantic validity would be lost if the sound were trimmed before the sound event ends, for example an animal vocalization such as a dog bark. random_state : int, RandomState instance or None, optional (default=None) If int, random_state is the seed used by the random number generator; If RandomState instance, random_state is the random number generator; If None, the random number generator is the RandomState instance used by np.random. Note that if the random state is passed as a RandomState instance, it is passed by reference, not value. This will lead to the Scaper object advancing the state of the random state object if you use it elsewhere. ''' # Duration must be a positive real number if np.isrealobj(duration) and duration > 0: self.duration = duration else: raise ScaperError('Duration must be a positive real value') # Initialize parameters self.sr = 44100 self.ref_db = -12 self.n_channels = 1 self.fade_in_len = 0.01 # 10 ms self.fade_out_len = 0.01 # 10 ms # Start with empty specifications self.fg_spec = [] self.bg_spec = [] # Validate paths and set expanded_fg_path = os.path.expanduser(fg_path) expanded_bg_path = os.path.expanduser(bg_path) _validate_folder_path(expanded_fg_path) _validate_folder_path(expanded_bg_path) self.fg_path = expanded_fg_path self.bg_path = expanded_bg_path # Populate label lists from folder paths self.fg_labels = [] self.bg_labels = [] _populate_label_list(self.fg_path, self.fg_labels) _populate_label_list(self.bg_path, self.bg_labels) # Copy list of protected labels self.protected_labels = protected_labels[:] # Get random number generator self.random_state = _check_random_state(random_state) def reset_fg_event_spec(self): ''' Resets the foreground event specification to be an empty list as it is when the Scaper object is initialized in the first place. This allows the same Scaper object to be used over and over again to generate new soundscapes with the same underlying settings (e.g. `ref_db`, `num_channels`, and so on.) See Also -------- Scaper.reset_bg_event_spec : Same functionality but resets the background event specification instead of the foreground specification. ''' self.fg_spec = [] def reset_bg_event_spec(self): ''' Resets the background event specification to be an empty list as it is when the Scaper object is initialized in the first place. This allows the same Scaper object to be used over and over again to generate new soundscapes with the same underlying settings (e.g. `ref_db`, `num_channels`, and so on.) See Also -------- Scaper.reset_fg_event_spec : Same functionality but resets the foreground event specification instead of the foreground specification. ''' self.bg_spec = [] def set_random_state(self, random_state): ''' Allows the user to set the random state after creating the Scaper object. Parameters ---------- random_state : int, RandomState instance or None, optional (default=None) If int, random_state is the seed used by the random number generator; If RandomState instance, random_state is the random number generator; If None, the random number generator is the RandomState instance used by np.random. ''' self.random_state = _check_random_state(random_state) def add_background(self, label, source_file, source_time): ''' Add a background recording to the background specification. The background duration will be equal to the duration of the soundscape ``Scaper.duration`` specified when initializing the Scaper object. If the source file is shorter than this duration then it will be concatenated to itself as many times as necessary to produce the specified duration when calling ``Scaper.generate``. Parameters ---------- label : tuple Specifies the label of the background. See Notes below for the expected format of this tuple and the allowed values. NOTE: The label specified by this tuple must match one of the labels in the Scaper's background label list ``Scaper.bg_labels``. Furthermore, if ``source_file`` is specified using "const" (see Notes), then ``label`` must also be specified using "const" and its value (see Notes) must match the source file's parent folder's name. source_file : tuple Specifies the audio file to use as the source. See Notes below for the expected format of this tuple and the allowed values. NOTE: If ``source_file`` is specified using "const" (see Notes), then ``label`` must also be specified using "const" and its value (see Notes) must match the source file's parent folder's name. source_time : tuple Specifies the desired start time in the source file. See Notes below for the expected format of this tuple and the allowed values. NOTE: the source time specified by this tuple should be equal to or smaller than ``<source file duration> - <soundscape duration>``. Larger values will be automatically changed to fulfill this requirement when calling ``Scaper.generate``. Notes ----- Each parameter of this function is set by passing a distribution tuple, whose first item is always the distribution name and subsequent items are distribution specific. The supported distribution tuples are: * ``("const", value)`` : a constant, given by ``value``. * ``("choose", valuelist)`` : choose a value from ``valuelist`` at random (uniformly). The ``label`` and ``source_file`` parameters also support providing an empty ``valuelist`` i.e. ``("choose", [])``, in which case the value will be chosen at random from all available labels or files as determined automatically by Scaper by examining the file structure of ``bg_path`` provided during initialization. * ``("choose_weighted", valuelist, probabilities)``: choose a value from ``valuelist`` via weighted sampling, where the probability of sampling ``valuelist[i]`` is given by ``probabilities[i]``. The ``probabilities`` list must contain a valid probability distribution, i.e., all values must be in the range [0, 1] and sum to one. * ``("uniform", min_value, max_value)`` : sample a random value from a uniform distribution between ``min_value`` and ``max_value``. * ``("normal", mean, stddev)`` : sample a random value from a normal distribution defined by its mean ``mean`` and standard deviation ``stddev``. * ``("truncnorm", mean, stddev, min, max)``: sapmle a random value from a truncated normal distribution defined by its mean ``mean``, standard deviation ``stddev``, minimum value ``min`` and maximum value ``max``. IMPORTANT: not all parameters support all distribution tuples. In particular, ``label`` and ``source_file`` only support ``"const"``, ``"choose"`` and ``choose_weighted``, whereas ``source_time`` supports all distribution tuples. As noted above, only ``label`` and ``source_file`` support providing an empty ``valuelist`` with ``"choose"``. ''' # These values are fixed for the background sound event_time = ("const", 0) event_duration = ("const", self.duration) snr = ("const", 0) role = 'background' pitch_shift = None time_stretch = None # Validate parameter format and values _validate_event(label, source_file, source_time, event_time, event_duration, snr, self.bg_labels, None, None) # Create background sound event bg_event = EventSpec(label=label, source_file=source_file, source_time=source_time, event_time=event_time, event_duration=event_duration, snr=snr, role=role, pitch_shift=pitch_shift, time_stretch=time_stretch) # Add event to background spec self.bg_spec.append(bg_event) def add_event(self, label, source_file, source_time, event_time, event_duration, snr, pitch_shift, time_stretch): ''' Add a foreground sound event to the foreground specification. Parameters ---------- label : tuple Specifies the label of the sound event. See Notes below for the expected format of this tuple and the allowed values. NOTE: The label specified by this tuple must match one of the labels in the Scaper's foreground label list ``Scaper.fg_labels``. Furthermore, if ``source_file`` is specified using "const" (see Notes), then ``label`` must also be specified using "const" and its ``value `` (see Notes) must match the source file's parent folder's name. source_file : tuple Specifies the audio file to use as the source. See Notes below for the expected format of this tuple and the allowed values. NOTE: If ``source_file`` is specified using "const" (see Notes), then ``label`` must also be specified using "const" and its ``value`` (see Notes) must match the source file's parent folder's name. source_time : tuple Specifies the desired start time in the source file. See Notes below for the expected format of this tuple and the allowed values. NOTE: the source time specified by this tuple should be equal to or smaller than ``<source file duration> - event_duration``. Larger values will be automatically changed to fulfill this requirement when calling ``Scaper.generate``. event_time : tuple Specifies the desired start time of the event in the soundscape. See Notes below for the expected format of this tuple and the allowed values. NOTE: The value specified by this tuple should be equal to or smaller than ``<soundscapes duration> - event_duration``, and larger values will be automatically changed to fulfill this requirement when calling ``Scaper.generate``. event_duration : tuple Specifies the desired duration of the event. See Notes below for the expected format of this tuple and the allowed values. NOTE: The value specified by this tuple should be equal to or smaller than the source file's duration, and larger values will be automatically changed to fulfill this requirement when calling ``Scaper.generate``. snr : tuple Specifies the desired signal to noise ratio (SNR) between the event and the background. See Notes below for the expected format of this tuple and the allowed values. pitch_shift : tuple Specifies the number of semitones to shift the event by. None means no pitch shift. time_stretch: tuple Specifies the time stretch factor (value>1 will make it slower and longer, value<1 will makes it faster and shorter). Notes ----- Each parameter of this function is set by passing a distribution tuple, whose first item is always the distribution name and subsequent items are distribution specific. The supported distribution tuples are: * ``("const", value)`` : a constant, given by ``value``. * ``("choose", valuelist)`` : choose a value from ``valuelist`` at random (uniformly). The ``label`` and ``source_file`` parameters also support providing an empty ``valuelist`` i.e. ``("choose", [])``, in which case the value will be chosen at random from all available labels or source files as determined automatically by Scaper by examining the file structure of ``fg_path`` provided during initialization. * ``("choose_weighted", valuelist, probabilities)``: choose a value from ``valuelist`` via weighted sampling, where the probability of sampling ``valuelist[i]`` is given by ``probabilities[i]``. The ``probabilities`` list must contain a valid probability distribution, i.e., all values must be in the range [0, 1] and sum to one. * ``("uniform", min_value, max_value)`` : sample a random value from a uniform distribution between ``min_value`` and ``max_value`` (including ``max_value``). * ``("normal", mean, stddev)`` : sample a random value from a normal distribution defined by its mean ``mean`` and standard deviation ``stddev``. * ``("truncnorm", mean, stddev, min, max)``: sapmle a random value from a truncated normal distribution defined by its mean ``mean``, standard deviation ``stddev``, minimum value ``min`` and maximum value ``max``. IMPORTANT: not all parameters support all distribution tuples. In particular, ``label`` and ``source_file`` only support ``"const"``, ``"choose"`` and ``"choose_weighted"``, whereas the remaining parameters support all distribution tuples. As noted above, only ``label`` and ``source_file`` support providing an empty ``valuelist`` with ``"choose"``. See Also -------- _validate_event : Check that event parameter values are valid. Scaper.generate : Generate a soundscape based on the current specification and save to disk as both an audio file and a JAMS file describing the soundscape. ''' # SAFETY CHECKS _validate_event(label, source_file, source_time, event_time, event_duration, snr, self.fg_labels, pitch_shift, time_stretch) # Create event event = EventSpec(label=label, source_file=source_file, source_time=source_time, event_time=event_time, event_duration=event_duration, snr=snr, role='foreground', pitch_shift=pitch_shift, time_stretch=time_stretch) # Add event to foreground specification self.fg_spec.append(event) def _instantiate_event(self, event, isbackground=False, allow_repeated_label=True, allow_repeated_source=True, used_labels=[], used_source_files=[], disable_instantiation_warnings=False): ''' Instantiate an event specification. Given an event specification containing distribution tuples, instantiate the event, i.e. samples values for the label, source_file, source_time, event_time, event_duration and snr from their respective distribution tuples, and return the sampled values in as a new event specification. Parameters ---------- event : EventSpec Event specification containing distribution tuples. isbackground : bool Flag indicating whether the event to instantiate is a background event or not (False implies it is a foreground event). allow_repeated_label : bool When True (default) any label can be used, including a label that has already been used for another event. When False, only a label that is not already in ``used_labels`` can be selected. allow_repeated_source : bool When True (default) any source file matching the selected label can be used, including a source file that has already been used for another event. When False, only a source file that is not already in ``used_source_files`` can be selected. used_labels : list List labels that have already been used in the current soundscape instantiation. The label selected for instantiating the event will be appended to this list unless its already in it. used_source_files : list List of full paths to source files that have already been used in the current soundscape instantiation. The source file selected for instantiating the event will be appended to this list unless its already in it. disable_instantiation_warnings : bool When True (default is False), warnings stemming from event instantiation (primarily about automatic duration adjustments) are disabled. Not recommended other than for testing purposes. Returns ------- instantiated_event : EventSpec Event specification containing values sampled from the distribution tuples of the input event specification. Raises ------ ScaperError If allow_repeated_source is False and there is no valid source file to select. ''' # set paths and labels depending on whether its a foreground/background # event if isbackground: file_path = self.bg_path allowed_labels = self.bg_labels else: file_path = self.fg_path allowed_labels = self.fg_labels # determine label # special case: choose tuple with empty list if event.label[0] == "choose" and not event.label[1]: label_tuple = list(event.label) label_tuple[1] = allowed_labels label_tuple = tuple(label_tuple) else: label_tuple = event.label label = _get_value_from_dist(label_tuple, self.random_state) # Make sure we can use this label if (not allow_repeated_label) and (label in used_labels): if (len(allowed_labels) == len(used_labels) or label_tuple[0] == "const"): raise ScaperError( "Cannot instantiate event {:s}: all available labels " "have already been used and " "allow_repeated_label=False.".format(label)) else: while label in used_labels: label = _get_value_from_dist(label_tuple, self.random_state) # Update the used labels list if label not in used_labels: used_labels.append(label) # determine source file # special case: choose tuple with empty list if event.source_file[0] == "choose" and not event.source_file[1]: source_files = _get_sorted_files(os.path.join(file_path, label)) source_file_tuple = list(event.source_file) source_file_tuple[1] = source_files source_file_tuple = tuple(source_file_tuple) else: source_file_tuple = event.source_file source_file = _get_value_from_dist(source_file_tuple, self.random_state) # Make sure we can use this source file if (not allow_repeated_source) and (source_file in used_source_files): source_files = _get_sorted_files(os.path.join(file_path, label)) if (len(source_files) == len(used_source_files) or source_file_tuple[0] == "const"): raise ScaperError( "Cannot instantiate event {:s}: all available source " "files have already been used and " "allow_repeated_source=False.".format(label)) else: while source_file in used_source_files: source_file = _get_value_from_dist(source_file_tuple, self.random_state) # Update the used source files list if source_file not in used_source_files: used_source_files.append(source_file) # Get the duration of the source audio file source_duration = soundfile.info(source_file).duration # If this is a background event, the event duration is the # duration of the soundscape. if isbackground: event_duration = self.duration # If the foreground event's label is in the protected list, use the # source file's duration without modification. elif label in self.protected_labels: event_duration = source_duration else: # determine event duration # For background events the duration is fixed to self.duration # (which must be > 0), but for foreground events it could # potentially be non-positive, hence the loop. event_duration = -np.Inf while event_duration <= 0: event_duration = _get_value_from_dist( event.event_duration, self.random_state ) # Check if chosen event duration is longer than the duration of the # selected source file, if so adjust the event duration. if (event_duration > source_duration): old_duration = event_duration # for warning event_duration = source_duration if not disable_instantiation_warnings: warnings.warn( "{:s} event duration ({:.2f}) is greater that source " "duration ({:.2f}), changing to {:.2f}".format( label, old_duration, source_duration, event_duration), ScaperWarning) # Get time stretch value if event.time_stretch is None: time_stretch = None event_duration_stretched = event_duration else: time_stretch = -np.Inf while time_stretch <= 0: time_stretch = _get_value_from_dist( event.time_stretch, self.random_state ) # compute duration after stretching event_duration_stretched = event_duration * time_stretch # If the event duration is longer than the soundscape we can trim it # without losing validity (since the event will end when the soundscape # ends). if time_stretch is None: if event_duration > self.duration: old_duration = event_duration # for warning event_duration = self.duration if not disable_instantiation_warnings: warnings.warn( "{:s} event duration ({:.2f}) is greater than the " "soundscape duration ({:.2f}), changing to " "{:.2f}".format( label, old_duration, self.duration, self.duration), ScaperWarning) else: if event_duration_stretched > self.duration: old_duration = event_duration # for warning event_duration = self.duration / float(time_stretch) event_duration_stretched = self.duration if not disable_instantiation_warnings: warnings.warn( "{:s} event duration ({:.2f}) with stretch factor " "{:.2f} gives {:.2f} which is greater than the " "soundscape duration ({:.2f}), changing to " "{:.2f} ({:.2f} after time stretching)".format( label, old_duration, time_stretch, old_duration * time_stretch, self.duration, event_duration, event_duration_stretched), ScaperWarning) # Modify event.source_time so that sampling from the source time distribution # stays within the bounds of the audio file - event_duration. This allows users # to sample from anywhere in a source file without knowing the exact duration # of every source file. Only modify if label is not in protected labels. if label not in self.protected_labels: tuple_still_invalid = False modified_source_time, warn = _ensure_satisfiable_source_time_tuple( event.source_time, source_duration, event_duration ) # determine source time and also check again just in case (for normal dist). # if it happens again, just use the old method. source_time = -np.Inf while source_time < 0: source_time = _get_value_from_dist( modified_source_time, self.random_state) if source_time + event_duration > source_duration: source_time = max(0, source_duration - event_duration) warn = True tuple_still_invalid = True if warn and not disable_instantiation_warnings: old_source_time = ', '.join(map(str, event.source_time)) new_source_time = ', '.join(map(str, modified_source_time)) if not tuple_still_invalid: warnings.warn( "{:s} source time tuple ({:s}) could not be satisfied given " "source duration ({:.2f}) and event duration ({:.2f}), " "source time tuple changed to ({:s})".format( label, old_source_time, source_duration, event_duration, new_source_time), ScaperWarning) else: warnings.warn( "{:s} source time tuple ({:s}) could not be satisfied given " "source duration ({:.2f}) and event duration ({:.2f}), " "source time tuple changed to ({:s}) but was still not " "satisfiable, likely due to using 'normal' distribution with " "bounds too close to the start or end of the audio file".format( label, old_source_time, source_duration, event_duration, new_source_time), ScaperWarning) else: source_time = 0.0 # determine event time # for background events the event time is fixed to 0, but for # foreground events it's not. event_time = -np.Inf while event_time < 0: event_time = _get_value_from_dist( event.event_time, self.random_state ) # Make sure the selected event time + event duration are is not greater # than the total duration of the soundscape, if it is adjust the event # time. This means event duration takes precedence over the event # start time. if time_stretch is None: if event_time + event_duration > self.duration: old_event_time = event_time event_time = self.duration - event_duration if not disable_instantiation_warnings: warnings.warn( '{:s} event time ({:.2f}) is too great given event ' 'duration ({:.2f}) and soundscape duration ({:.2f}), ' 'changed to {:.2f}.'.format( label, old_event_time, event_duration, self.duration, event_time), ScaperWarning) else: if event_time + event_duration_stretched > self.duration: old_event_time = event_time event_time = self.duration - event_duration_stretched if not disable_instantiation_warnings: warnings.warn( '{:s} event time ({:.2f}) is too great given ' 'stretched event duration ({:.2f}) and soundscape ' 'duration ({:.2f}), changed to {:.2f}.'.format( label, old_event_time, event_duration_stretched, self.duration, event_time), ScaperWarning) # determine snr snr = _get_value_from_dist(event.snr, self.random_state) # get role (which can only take "foreground" or "background" and # is set internally, not by the user). role = event.role # determine pitch_shift if event.pitch_shift is not None: pitch_shift = _get_value_from_dist(event.pitch_shift, self.random_state) else: pitch_shift = None # pack up instantiated values in an EventSpec instantiated_event = EventSpec(label=label, source_file=source_file, source_time=source_time, event_time=event_time, event_duration=event_duration, snr=snr, role=role, pitch_shift=pitch_shift, time_stretch=time_stretch) # Return return instantiated_event def _instantiate(self, allow_repeated_label=True, allow_repeated_source=True, reverb=None, disable_instantiation_warnings=False): ''' Instantiate a specific soundscape in JAMS format based on the current specification. Any non-deterministic event values (i.e. distribution tuples) will be sampled randomly from based on the distribution parameters. Parameters ---------- allow_repeated_label : bool When True (default) the same label can be used more than once in a soundscape instantiation. When False every label can only be used once. allow_repeated_source : bool When True (default) the same source file can be used more than once in a soundscape instantiation. When False every source file can only be used once. reverb : float or None Has no effect on this function other than being documented in the instantiated annotation's sandbox. Passed by ``Scaper.generate``. disable_instantiation_warnings : bool When True (default is False), warnings stemming from event instantiation (primarily about automatic duration adjustments) are disabled. Not recommended other than for testing purposes. Returns ------- jam : JAMS object A JAMS object containing a scaper annotation representing the instantiated soundscape. See Also -------- Scaper.generate ''' jam = jams.JAMS() ann = jams.Annotation(namespace='scaper') # Set annotation duration (might be changed later due to cropping) ann.duration = self.duration # INSTANTIATE BACKGROUND AND FOREGROUND EVENTS AND ADD TO ANNOTATION # NOTE: logic for instantiating bg and fg events is NOT the same. # Add background sounds bg_labels = [] bg_source_files = [] for event in self.bg_spec: value = self._instantiate_event( event, isbackground=True, allow_repeated_label=allow_repeated_label, allow_repeated_source=allow_repeated_source, used_labels=bg_labels, used_source_files=bg_source_files, disable_instantiation_warnings=disable_instantiation_warnings) # Note: add_background doesn't allow to set a time_stretch, i.e. # it's hardcoded to time_stretch=None, so we don't need to check # if value.time_stretch is not None, since it always will be. ann.append(time=value.event_time, duration=value.event_duration, value=value._asdict(), confidence=1.0) # Add foreground events fg_labels = [] fg_source_files = [] for event in self.fg_spec: value = self._instantiate_event( event, isbackground=False, allow_repeated_label=allow_repeated_label, allow_repeated_source=allow_repeated_source, used_labels=fg_labels, used_source_files=fg_source_files, disable_instantiation_warnings=disable_instantiation_warnings) if value.time_stretch is not None: event_duration_stretched = ( value.event_duration * value.time_stretch) else: event_duration_stretched = value.event_duration ann.append(time=value.event_time, duration=event_duration_stretched, value=value._asdict(), confidence=1.0) # Compute max polyphony poly = max_polyphony(ann) # Compute the number of foreground events n_events = len(self.fg_spec) # Compute gini gini = polyphony_gini(ann) # Add specs and other info to sandbox ann.sandbox.scaper = jams.Sandbox( duration=self.duration, original_duration=self.duration, fg_path=self.fg_path, bg_path=self.bg_path, fg_spec=self.fg_spec, bg_spec=self.bg_spec, fg_labels=self.fg_labels, bg_labels=self.bg_labels, protected_labels=self.protected_labels, sr=self.sr, ref_db=self.ref_db, n_channels=self.n_channels, fade_in_len=self.fade_in_len, fade_out_len=self.fade_out_len, n_events=n_events, polyphony_max=poly, polyphony_gini=gini, allow_repeated_label=allow_repeated_label, allow_repeated_source=allow_repeated_source, reverb=reverb, scaper_version=scaper_version, soundscape_audio_path=None, isolated_events_audio_path=[], # Initialize missing generate parameters audio_path=None, jams_path=None, fix_clipping=None, peak_normalization=None, save_isolated_events=None, isolated_events_path=None, disable_sox_warnings=None, no_audio=None, txt_path=None, txt_sep=None, disable_instantiation_warnings=None, peak_normalization_scale_factor=None, ref_db_change=None, ref_db_generated=None) # Add annotation to jams jam.annotations.append(ann) # Set jam metadata jam.file_metadata.duration = ann.duration # Return return jam def _generate_audio(self, audio_path, ann, reverb=None, fix_clipping=False, peak_normalization=False, quick_pitch_time=False, save_isolated_events=False, isolated_events_path=None, disable_sox_warnings=True): ''' Generate audio based on a scaper annotation and save to disk. Parameters ---------- audio_path : str Path for saving soundscape audio file. ann : jams.Annotation Annotation of the scaper namespace. reverb : float or None Amount of reverb to apply to the generated soundscape between 0 (no reverberation) and 1 (maximum reverberation). Use None (default) to prevent the soundscape from going through the reverb module at all. fix_clipping: bool When True (default=False), checks the soundscape audio for clipping (abs(sample) > 1). If so, the soundscape waveform is peak normalized, i.e., scaled such that max(abs(soundscape_audio)) = 1. The audio for each isolated event is also scaled accordingly. Note: this will change the actual value of `ref_db` in the generated audio. The scaling factor that was used is returned. peak_normalization : bool When True (default=False), normalize the generated soundscape audio such that max(abs(soundscape_audio)) = 1. The audio for each isolated event is also scaled accordingly. Note: this will change the actual value of `ref_db` in the generated audio. The scaling factor that was used is returned. quick_pitch_time : bool When True (default=False), time stretching and pitch shifting will be applied with `quick=True`. This is much faster but the resultant audio is generally of lower audio quality. save_isolated_events : bool If True, this will save the isolated foreground events and backgrounds in a directory adjacent to the generated soundscape mixture, or to the path defined by `isolated_events_path`. The audio of the isolated events sum up to the mixture if reverb is not applied. Isolated events can be found (by default) at `<audio_outfile parent folder>/<audio_outfile name>_events`. Isolated event file names follow the pattern: `<role><idx>_<label>`, where idx is the index of the isolated event in self.fg_spec or self.bg_spec (this allows events of the same label to be added more than once to the soundscape without breaking things). Role is "background" or "foreground". For example: `foreground0_siren.wav` or `background0_park.wav`. isolated_events_path : str Path to folder for saving isolated events. If None, defaults to `<audio_path parent folder>/<audio_path name>_events`. disable_sox_warnings : bool When True (default), warnings from the pysox module are suppressed unless their level is ``'CRITICAL'``. Returns ------- soundscape_audio : np.ndarray The audio samples of the generated soundscape event_audio_list: list A list of np.ndarrays containing the audio samples of every individual background and foreground sound event. Events are listed in the same order in which they appear in the jams annotations data list, and can be matched with: `for obs, event_audio in zip(ann.data, event_audio_list): ...`. scale_factor : float If peak_normalization is True, or fix_clipping is True and the soundscape audio needs to be scaled to avoid clipping, scale_factor is the value used to scale the soundscape audio and the audio of the isolated events. Otherwise will return 1.0. ref_db_change : float The change (in dB) to the soundscape audio's ref_db if peak normalization is applied to fix clipping or because the user specified it. Otherwise will return 0. Raises ------ ScaperError If annotation is not of the scpaper namespace. See Also -------- Scaper.generate ''' if ann.namespace != 'scaper': raise ScaperError( 'Annotation namespace must be scaper, found: {:s}'.format( ann.namespace)) # disable sox warnings if disable_sox_warnings: temp_logging_level = 'CRITICAL' # only critical messages please else: temp_logging_level = logging.getLogger().level # List for storing all generated audio (one array for every event) soundscape_audio = None event_audio_list = [] scale_factor = 1.0 ref_db_change = 0 with _set_temp_logging_level(temp_logging_level): isolated_events_audio_path = [] duration_in_samples = int(self.duration * self.sr) for i, e in enumerate(ann.data): if e.value['role'] == 'background': # Concatenate background if necessary. source_duration = soundfile.info(e.value['source_file']).duration ntiles = int( max(self.duration // source_duration + 1, 1)) # Create transformer tfm = sox.Transformer() # Ensure consistent sampling rate and channels # Need both a convert operation (to do the conversion), # and set_output_format (to have sox interpret the output # correctly). tfm.convert( samplerate=self.sr, n_channels=self.n_channels, bitdepth=None ) tfm.set_output_format( rate=self.sr, channels=self.n_channels ) # PROCESS BEFORE COMPUTING LUFS tmpfiles_internal = [] with _close_temp_files(tmpfiles_internal): # create internal tmpfile tmpfiles_internal.append( tempfile.NamedTemporaryFile( suffix='.wav', delete=False)) # read in background off disk, using start and stop # to only read the necessary audio event_sr = soundfile.info(e.value['source_file']).samplerate start = int(e.value['source_time'] * event_sr) stop = int((e.value['source_time'] + e.value['event_duration']) * event_sr) event_audio, event_sr = soundfile.read( e.value['source_file'], always_2d=True, start=start, stop=stop) # tile the background along the appropriate dimensions event_audio = np.tile(event_audio, (ntiles, 1)) event_audio = event_audio[:stop] event_audio = tfm.build_array( input_array=event_audio, sample_rate_in=event_sr ) event_audio = event_audio.reshape(-1, self.n_channels) # NOW compute LUFS bg_lufs = get_integrated_lufs(event_audio, self.sr) # Normalize background to reference DB. gain = self.ref_db - bg_lufs event_audio = np.exp(gain * np.log(10) / 20) * event_audio event_audio_list.append(event_audio[:duration_in_samples]) elif e.value['role'] == 'foreground': # Create transformer tfm = sox.Transformer() # Ensure consistent sampling rate and channels # Need both a convert operation (to do the conversion), # and set_output_format (to have sox interpret the output # correctly). tfm.convert( samplerate=self.sr, n_channels=self.n_channels, bitdepth=None ) tfm.set_output_format( rate=self.sr, channels=self.n_channels ) # Pitch shift if e.value['pitch_shift'] is not None: tfm.pitch(e.value['pitch_shift'], quick=quick_pitch_time) # Time stretch if e.value['time_stretch'] is not None: factor = 1.0 / float(e.value['time_stretch']) tfm.tempo(factor, audio_type='s', quick=quick_pitch_time) # PROCESS BEFORE COMPUTING LUFS tmpfiles_internal = [] with _close_temp_files(tmpfiles_internal): # create internal tmpfile tmpfiles_internal.append( tempfile.NamedTemporaryFile( suffix='.wav', delete=False)) # synthesize edited foreground sound event, # doing the trim via soundfile event_sr = soundfile.info(e.value['source_file']).samplerate start = int(e.value['source_time'] * event_sr) stop = int((e.value['source_time'] + e.value['event_duration']) * event_sr) event_audio, event_sr = soundfile.read( e.value['source_file'], always_2d=True, start=start, stop=stop) event_audio = tfm.build_array( input_array=event_audio, sample_rate_in=event_sr ) event_audio = event_audio.reshape(-1, self.n_channels) # NOW compute LUFS fg_lufs = get_integrated_lufs(event_audio, self.sr) # Normalize to specified SNR with respect to # background gain = self.ref_db + e.value['snr'] - fg_lufs event_audio = np.exp(gain * np.log(10) / 20) * event_audio # Apply short fade in and out # (avoid unnatural sound onsets/offsets) if self.fade_in_len > 0: fade_in_samples = int(self.fade_in_len * self.sr) fade_in_window = np.sin(np.linspace(0, np.pi / 2, fade_in_samples))[..., None] event_audio[:fade_in_samples] *= fade_in_window if self.fade_out_len > 0: fade_out_samples = int(self.fade_out_len * self.sr) fade_out_window = np.sin(np.linspace(np.pi / 2, 0, fade_out_samples))[..., None] event_audio[-fade_out_samples:] *= fade_out_window # Pad with silence before/after event to match the # soundscape duration prepad = int(self.sr * e.value['event_time']) postpad = max(0, duration_in_samples - (event_audio.shape[0] + prepad)) event_audio = np.pad(event_audio, ((prepad, postpad), (0, 0)), mode='constant', constant_values=(0, 0)) event_audio = event_audio[:duration_in_samples] event_audio_list.append(event_audio[:duration_in_samples]) else: raise ScaperError( 'Unsupported event role: {:s}'.format( e.value['role'])) # Finally combine all the files and optionally apply reverb. # If there are no events, throw a warning. if len(event_audio_list) == 0: warnings.warn( "No events to synthesize (silent soundscape), no audio " "generated.", ScaperWarning) else: # Sum all events to get soundscape audio soundscape_audio = sum(event_audio_list) # Check for clipping and fix [optional] max_sample = np.max(np.abs(soundscape_audio)) clipping = max_sample > 1 if clipping: warnings.warn('Soundscape audio is clipping!', ScaperWarning) if peak_normalization or (clipping and fix_clipping): # normalize soundscape audio and scale event audio soundscape_audio, event_audio_list, scale_factor = \ peak_normalize(soundscape_audio, event_audio_list) ref_db_change = 20 * np.log10(scale_factor) if clipping and fix_clipping: warnings.warn( 'Peak normalization applied to fix clipping with ' 'scale factor = {}. The actual ref_db of the ' 'generated soundscape audio will change by ' 'approximately {:.2f}dB with respect to the target ' 'ref_db of {})'.format( scale_factor, ref_db_change, self.ref_db), ScaperWarning) if scale_factor < 0.05: warnings.warn( 'Scale factor for peak normalization is extreme ' '(<0.05), event SNR values in the generated soundscape ' 'audio may not perfectly match their specified values.', ScaperWarning ) # Optionally apply reverb # NOTE: must apply AFTER peak normalization: applying reverb # to a clipping signal with sox and then normalizing doesn't # work as one would hope. if reverb is not None: tfm = sox.Transformer() tfm.reverb(reverberance=reverb * 100) soundscape_audio = tfm.build_array( input_array=soundscape_audio, sample_rate_in=self.sr, ) # Reshape to ensure data are 2d soundscape_audio = soundscape_audio.reshape(-1, self.n_channels) # Optionally save soundscape audio to disk if audio_path is not None: soundfile.write(audio_path, soundscape_audio, self.sr, subtype='PCM_32') # Optionally save isolated events to disk if save_isolated_events: base, ext = os.path.splitext(audio_path) if isolated_events_path is None: event_folder = '{:s}_events'.format(base) else: event_folder = isolated_events_path os.makedirs(event_folder, exist_ok=True) iso_idx = 0 role_counter = {'background': 0, 'foreground': 0} for i, e in enumerate(ann.data): _role_count = role_counter[e.value['role']] event_audio_path = os.path.join( event_folder, '{:s}{:d}_{:s}{:s}'.format( e.value['role'], _role_count, e.value['label'], ext)) role_counter[e.value['role']] += 1 soundfile.write(event_audio_path, event_audio_list[iso_idx], self.sr, subtype='PCM_32') isolated_events_audio_path.append(event_audio_path) iso_idx += 1 # TODO what do we do in this case? for now throw a warning if reverb is not None: warnings.warn( "Reverb is on and save_isolated_events is True. Reverberation " "is applied to the mixture but not output " "source files. In this case the sum of the " "audio of the isolated events will not add up to the " "mixture", ScaperWarning) # Document output paths # TODO: this is redundant with audio_path and isolated_events_path that # are also stored in ann.sandbox.scaper. For now we're keeping these # here for now for backwards compatibility e.g. with FUSS. Eventually # we should remove these two lines and consolidate how/where JAMS # metadata is stored (cf. generate() and generate_from_jams()). ann.sandbox.scaper.soundscape_audio_path = audio_path ann.sandbox.scaper.isolated_events_audio_path = isolated_events_audio_path # Return audio for in-memory processing return soundscape_audio, event_audio_list, scale_factor, ref_db_change def generate(self, audio_path=None, jams_path=None, allow_repeated_label=True, allow_repeated_source=True, reverb=None, fix_clipping=False, peak_normalization=False, quick_pitch_time=False, save_isolated_events=False, isolated_events_path=None, disable_sox_warnings=True, no_audio=False, txt_path=None, txt_sep='\t', disable_instantiation_warnings=False): """ Generate a soundscape based on the current specification and return as an audio file, a JAMS annotation, a simplified annotation list, and a list containing the audio samples of each individual background and foreground event. If output paths are provided, these objects will also be saved to disk. Parameters ---------- audio_path : str Path for saving soundscape audio to disk. If None, does not save audio to disk. jams_path : str Path for saving soundscape jams annotation to disk. If None, does not save JAMS to disk. allow_repeated_label : bool When True (default) the same label can be used more than once in a soundscape instantiation. When False every label can only be used once. allow_repeated_source : bool When True (default) the same source file can be used more than once in a soundscape instantiation. When False every source file can only be used once. reverb : float or None Amount of reverb to apply to the generated soundscape between 0 (no reverberation) and 1 (maximum reverberation). Use None (default) to prevent the soundscape from going through the reverb module at all. fix_clipping: bool When True (default=False), checks the soundscape audio for clipping (abs(sample) > 1). If so, the soundscape waveform is peak normalized, i.e., scaled such that max(abs(soundscape_audio)) = 1. The audio for each isolated event is also scaled accordingly. Note: this will change the actual value of `ref_db` in the generated audio. The updated `ref_db` value will be stored in the JAMS annotation. The SNR of foreground events with respect to the background is unaffected except when extreme scaling is required to prevent clipping. peak_normalization : bool When True (default=False), normalize the generated soundscape audio such that max(abs(soundscape_audio)) = 1. The audio for each isolated event is also scaled accordingly. Note: this will change the actual value of `ref_db` in the generated audio. The updated `ref_db` value will be stored in the JAMS annotation. The SNR of foreground events with respect to the background is unaffected except when extreme scaling is required to achieve peak normalization. quick_pitch_time : bool When True (default=False), time stretching and pitch shifting will be applied with `quick=True`. This is much faster but the resultant audio is generally of lower audio quality. save_isolated_events : bool If True, this will save the isolated foreground events and backgrounds in a directory adjacent to the generated soundscape mixture, or to the path defined by `isolated_events_path`. The audio of the isolated events sum up to the mixture if reverb is not applied. Isolated events can be found (by default) at `<audio_outfile parent folder>/<audio_outfile name>_events`. Isolated event file names follow the pattern: `<role><idx>_<label>`, where count is the index of the isolated event in self.fg_spec or self.bg_spec (this allows events of the same label to be added more than once to the soundscape without breaking things). Role is "background" or "foreground". For example: `foreground0_siren.wav` or `background0_park.wav`. isolated_events_path : str Path to folder for saving isolated events. If None, defaults to `<audio_path parent folder>/<audio_path name>_events`. Only relevant if save_isolated_events=True. disable_sox_warnings : bool When True (default), warnings from the pysox module are suppressed unless their level is ``'CRITICAL'``. If you're experiencing issues related to audio I/O setting this parameter to False may help with debugging. no_audio : bool If True this function will only generates a JAMS file but will not generate any audio (neither in memory nor saved to disk). Useful for efficiently generating a large number of soundscape JAMS for later synthesis via `generate_from_jams()`. txt_path: str or None Path for saving a simplified annotation in a space separated format [onset offset label] where onset and offset are in seconds. Good for loading labels in e.g. Audacity. If None, does not save txt annotation to disk. txt_sep: str The separator to use when saving a simplified annotation as a text file (default is tab for compatibility with Audacity label files). Only relevant if txt_path is not None. disable_instantiation_warnings : bool When True (default is False), warnings stemming from event instantiation (primarily about automatic duration adjustments) are disabled. Not recommended other than for testing purposes. Returns ------- soundscape_audio : np.ndarray The audio samples of the generated soundscape. Returns None if no_audio=True. soundscape_jam: jams.JAMS The JAMS object containing the full soundscape annotation. annotation_list : list A simplified annotation in a space-separated format [onset offset label] where onset and offset are in seconds. event_audio_list: list A list of np.ndarrays containing the audio samples of every individual background and foreground sound event. Events are listed in the same order in which they appear in the jams annotations data list, and can be matched with: `for obs, event_audio in zip(ann.data, event_audio_list): ...`. Raises ------ ScaperError If the reverb parameter is passed an invalid value. See Also -------- Scaper.generate_from_jams Scaper._instantiate Scaper._generate_audio """ # Check parameter validity if reverb is not None: if not (0 <= reverb <= 1): raise ScaperError( 'Invalid value for reverb: must be in range [0, 1] or ' 'None.') # Create specific instance of a soundscape based on the spec soundscape_jam = self._instantiate( allow_repeated_label=allow_repeated_label, allow_repeated_source=allow_repeated_source, reverb=reverb, disable_instantiation_warnings=disable_instantiation_warnings) ann = soundscape_jam.annotations.search(namespace='scaper')[0] soundscape_audio, event_audio_list = None, None # Generate the audio and save to disk scale_factor = 1.0 ref_db_change = 0 if not no_audio: soundscape_audio, event_audio_list, scale_factor, ref_db_change = \ self._generate_audio(audio_path, ann, reverb=reverb, save_isolated_events=save_isolated_events, isolated_events_path=isolated_events_path, disable_sox_warnings=disable_sox_warnings, fix_clipping=fix_clipping, peak_normalization=peak_normalization, quick_pitch_time=quick_pitch_time) # TODO: Stick to heavy handed overwriting for now, in the future we # should consolidate this with what happens inside _instantiate(). ann.sandbox.scaper.audio_path = audio_path ann.sandbox.scaper.jams_path = jams_path ann.sandbox.scaper.allow_repeated_label = allow_repeated_label ann.sandbox.scaper.allow_repeated_source = allow_repeated_source ann.sandbox.scaper.reverb = reverb ann.sandbox.scaper.fix_clipping = fix_clipping ann.sandbox.scaper.peak_normalization = peak_normalization ann.sandbox.scaper.quick_pitch_time = quick_pitch_time ann.sandbox.scaper.save_isolated_events = save_isolated_events ann.sandbox.scaper.isolated_events_path = isolated_events_path ann.sandbox.scaper.disable_sox_warnings = disable_sox_warnings ann.sandbox.scaper.no_audio = no_audio ann.sandbox.scaper.txt_path = txt_path ann.sandbox.scaper.txt_sep = txt_sep ann.sandbox.scaper.disable_instantiation_warnings = disable_instantiation_warnings ann.sandbox.scaper.peak_normalization_scale_factor = scale_factor ann.sandbox.scaper.ref_db_change = ref_db_change ann.sandbox.scaper.ref_db_generated = self.ref_db + ref_db_change # Save JAMS to disk too if jams_path is not None: soundscape_jam.save(jams_path) # Create annotation list annotation_list = [] for obs in ann.data: if obs.value['role'] == 'foreground': annotation_list.append( [obs.time, obs.time + obs.duration, obs.value['label']]) if txt_path is not None: with open(txt_path, 'w') as csv_file: writer = csv.writer(csv_file, delimiter=txt_sep) writer.writerows(annotation_list) # Return return soundscape_audio, soundscape_jam, annotation_list, event_audio_list
106,280
43.712242
111
py
scaper
scaper-master/scaper/scaper_exceptions.py
#!/usr/bin/env python # -*- encoding: utf-8 -*- # CREATED: 10/11/16 6:07 PM by Justin Salamon <justin.salamon@nyu.edu> '''Exception classes for Scaper''' class ScaperError(Exception): '''The root Scaper exception class''' pass
239
19
70
py
scaper
scaper-master/scaper/audio.py
# CREATED: 4/23/17 15:37 by Justin Salamon <justin.salamon@nyu.edu> import numpy as np import pyloudnorm import soundfile from .scaper_exceptions import ScaperError def get_integrated_lufs(audio_array, samplerate, min_duration=0.5, filter_class='K-weighting', block_size=0.400): """ Returns the integrated LUFS for a numpy array containing audio samples. For files shorter than 400 ms pyloudnorm throws an error. To avoid this, files shorter than min_duration (by default 500 ms) are self-concatenated until min_duration is reached and the LUFS value is computed for the concatenated file. Parameters ---------- audio_array : np.ndarray numpy array containing samples or path to audio file for computing LUFS samplerate : int Sample rate of audio, for computing duration min_duration : float Minimum required duration for computing LUFS value. Files shorter than this are self-concatenated until their duration reaches this value for the purpose of computing the integrated LUFS. Caution: if you set min_duration < 0.4, a constant LUFS value of -70.0 will be returned for all files shorter than 400 ms. filter_class : str Class of weighting filter used. - 'K-weighting' (default) - 'Fenton/Lee 1' - 'Fenton/Lee 2' - 'Dash et al.' block_size : float Gating block size in seconds. Defaults to 0.400. Returns ------- loudness Loudness in terms of LUFS """ duration = audio_array.shape[0] / float(samplerate) if duration < min_duration: ntiles = int(np.ceil(min_duration / duration)) audio_array = np.tile(audio_array, (ntiles, 1)) meter = pyloudnorm.Meter( samplerate, filter_class=filter_class, block_size=block_size ) loudness = meter.integrated_loudness(audio_array) # silent audio gives -inf, so need to put a lower bound. loudness = max(loudness, -70) return loudness def match_sample_length(audio_path, duration_in_samples): ''' Takes a path to an audio file and a duration defined in samples. The audio is loaded from the specifid audio_path and padded or trimmed such that it matches the duration_in_samples. The modified audio is then saved back to audio_path. This ensures that the durations match exactly. If the audio needed to be padded, it is padded with zeros to the end of the audio file. If the audio needs to be trimmed, the function will trim samples from the end of the audio file. The sample rate of the saved audio is the same as the sample rate of the input file. Parameters ---------- audio_path : str Path to the audio file that will be modified. duration_in_samples : int Duration that the audio will be padded or trimmed to. ''' if duration_in_samples <= 0: raise ScaperError( 'Duration in samples must be > 0.') if not isinstance(duration_in_samples, int): raise ScaperError( 'Duration in samples must be an integer.') audio, sr = soundfile.read(audio_path) audio_info = soundfile.info(audio_path) current_duration = audio.shape[0] if duration_in_samples < current_duration: audio = audio[:duration_in_samples] elif duration_in_samples > current_duration: n_pad = duration_in_samples - current_duration pad_width = [(0, 0) for _ in range(len(audio.shape))] pad_width[0] = (0, n_pad) audio = np.pad(audio, pad_width, 'constant') soundfile.write(audio_path, audio, sr, subtype=audio_info.subtype, format=audio_info.format) def peak_normalize(soundscape_audio, event_audio_list): """ Compute the scale factor required to peak normalize the audio such that max(abs(soundscape_audio)) = 1. Parameters ---------- soundscape_audio : np.ndarray The soudnscape audio. event_audio_list : list List of np.ndarrays containing the audio samples of each isolated foreground event. Returns ------- scaled_soundscape_audio : np.ndarray The peak normalized soundscape audio. scaled_event_audio_list : list List of np.ndarrays containing the scaled audio samples of each isolated foreground event. All events are scaled by scale_factor. scale_factor : float The scale factor used to peak normalize the soundscape audio. """ eps = 1e-10 max_sample = np.max(np.abs(soundscape_audio)) scale_factor = 1.0 / (max_sample + eps) # scale the event audio and the soundscape audio: scaled_soundscape_audio = soundscape_audio * scale_factor scaled_event_audio_list = [] for event_audio in event_audio_list: scaled_event_audio_list.append(event_audio * scale_factor) return scaled_soundscape_audio, scaled_event_audio_list, scale_factor
4,982
35.108696
85
py
scaper
scaper-master/scaper/version.py
#!/usr/bin/env python # -*- coding: utf-8 -*- """Version info""" short_version = '1.6' version = '1.6.5'
106
14.285714
23
py
scaper
scaper-master/scaper/util.py
# CREATED: 10/14/16 12:35 PM by Justin Salamon <justin.salamon@nyu.edu> ''' Utility functions ================= ''' from contextlib import contextmanager import logging import os import glob from .scaper_exceptions import ScaperError import warnings from .scaper_warnings import ScaperWarning import scipy import numpy as np import numbers from copy import deepcopy @contextmanager def _close_temp_files(tmpfiles): ''' Utility function for creating a context and closing all temporary files once the context is exited. For correct functionality, all temporary file handles created inside the context must be appended to the ```tmpfiles``` list. Parameters ---------- tmpfiles : list List of temporary file handles ''' def _close(): for t in tmpfiles: try: t.close() os.unlink(t.name) except: pass try: yield except: _close() raise _close() @contextmanager def _set_temp_logging_level(level): ''' Utility function for temporarily changing the logging level using contexts. Parameters ---------- level : str or int The desired temporary logging level. For allowed values see: https://docs.python.org/2/library/logging.html#logging-levels ''' logger = logging.getLogger() current_level = logger.level logger.setLevel(level) yield logger.setLevel(current_level) def _get_sorted_files(folder_path): ''' Return a list of absolute paths to all valid files contained within the folder specified by ```folder_path```. Parameters ---------- folder_path : str Path to the folder to scan for files. Returns ------- files : list List of absolute paths to all valid files contained within ```folder_path```. ''' # Ensure path points to valid folder _validate_folder_path(folder_path) # Get folder contents and filter for valid files # Note, we sort the list to ensure consistent behavior across operating # systems. files = sorted(glob.glob(os.path.join(folder_path, "*"))) files = [f for f in files if os.path.isfile(f)] return files def _validate_folder_path(folder_path): ''' Validate that a provided path points to a valid folder. Parameters ---------- folder_path : str Path to a folder. Raises ------ ScaperError If ```folder_path``` does not point to a valid folder. ''' if not os.path.isdir(folder_path): raise ScaperError( 'Folder path "{:s}" does not point to a valid folder'.format( str(folder_path))) def _populate_label_list(folder_path, label_list): ''' Given a path to a folder and a list, add the names of all subfolders contained in this folder (excluding folders whose name starts with '.') to the provided list. This is used in scaper to populate the lists of valid foreground and background labels, which are determined by the names of the folders contained in ```fg_path`` and ```bg_path``` provided during initialization. Parameters ---------- folder_path : str Path to a folder label_list : list List to which label (subfolder) names will be added. See Also -------- _validate_folder_path : Validate that a provided path points to a valid folder. ''' # Make sure folder path is valid _validate_folder_path(folder_path) folder_names = os.listdir(folder_path) for fname in folder_names: if (os.path.isdir(os.path.join(folder_path, fname)) and fname[0] != '.'): label_list.append(fname) # ensure consistent ordering of labels label_list.sort() def _check_random_state(seed): """Turn seed into a np.random.RandomState instance Parameters ---------- seed : None | int | instance of RandomState If seed is None, return the RandomState singleton used by np.random. If seed is an int, return a new RandomState instance seeded with seed. If seed is already a RandomState instance, return it. Otherwise raise ValueError. """ if seed is None or seed is np.random: return np.random.mtrand._rand elif isinstance(seed, (numbers.Integral, np.integer, int)): return np.random.RandomState(seed) elif isinstance(seed, np.random.RandomState): return seed else: raise ValueError('%r cannot be used to seed a numpy.random.RandomState' ' instance' % seed) def _sample_const(item, random_state): ''' Return a value sampled from a constant distribution (just the item). Parameters ---------- item : any What to return random_state : mtrand.RandomState RandomState object used to sample from this distribution (ignored). This is here to match the other function specifications. Returns ------- value : any item, returned. ''' return item def _sample_uniform(minimum, maximum, random_state): ''' Return a random value sampled from a uniform distribution between ```minimum``` and ```maximum```. Parameters ---------- minimum : float Minimum of uniform distribution maximum : float Maximum of uniform distribution random_state : mtrand.RandomState RandomState object used to sample from this distribution. Returns ------- value : float A random value sampled from the uniform distribution defined by ```minimum```, ```maximum```. ''' return random_state.uniform(minimum, maximum) def _sample_normal(mu, sigma, random_state): ''' Return a random value sampled from a normal distribution with mean ```mu``` and standard deviation ```sigma```. Parameters ---------- mu : float The mean of the truncated normal distribution sig : float The standard deviation of the truncated normal distribution random_state : mtrand.RandomState RandomState object used to sample from this distribution. Returns ------- value : float A random value sampled from the normal distribution defined by ```mu```, ```sigma```. ''' return random_state.normal(mu, sigma) def _sample_choose(list_of_options, random_state): ''' Return a random item from ```list_of_options```, using random_state. If there are duplicates in ```list_of_options```, we remove them from the list before sampling an item from the list. Parameters ---------- list_of_options : list List of items to choose from. random_state : mtrand.RandomState RandomState object used to sample from this distribution. Returns ------- value : any A random item chosen from ```list_of_options```. ''' new_list_of_options = sorted(list(set(list_of_options))) if len(new_list_of_options) < len(list_of_options): warnings.warn( 'Removed duplicates from choose list. List length changed ' 'from {:d} to {:d}'.format(len(list_of_options), len(new_list_of_options)), ScaperWarning) index = random_state.randint(len(new_list_of_options)) return new_list_of_options[index] def _sample_choose_weighted(list_of_options, probabilities, random_state): ''' Return a random item from ```list_of_options``` using weighted sampling defined by ```probabilities```, using random_state. The number of items in ```list_of_options``` and ```probabilities``` must match, and the values in ```probabilities``` must be in the range [0, 1] and sum to 1. Unlike ```_sample_choose```, duplicates in ```list_of_options``` are not removed prior to sampling. Parameters ---------- list_of_options : list List of items to choose from. probabilities : list of floats List of probabilities corresponding to the elements in ```list_of_options```, such that the item in ```list_of_options[i]``` is chosen with probability ```probabilities[i]```. random_state : mtrand.RandomState RandomState object used to sample from this distribution. Returns ------- value : any A random item chosen from ```list_of_options```. ''' return random_state.choice(list_of_options, p=probabilities) def _sample_trunc_norm(mu, sigma, trunc_min, trunc_max, random_state): ''' Return a random value sampled from a truncated normal distribution with mean ```mu``` and standard deviation ```sigma``` whose values are limited between ```trunc_min``` and ```trunc_max```. Parameters ---------- mu : float The mean of the truncated normal distribution sig : float The standard deviation of the truncated normal distribution trunc_min : float The minimum value allowed for the distribution (lower boundary) trunc_max : float The maximum value allowed for the distribution (upper boundary) random_state : mtrand.RandomState RandomState object used to sample from this distribution. Returns ------- value : float A random value sampled from the truncated normal distribution defined by ```mu```, ```sigma```, ```trunc_min``` and ```trunc_max```. ''' # By default truncnorm expects a (lower boundary) and b (upper boundary) # values for a standard normal distribution (mu=0, sigma=1), so we need # to recompute a and b given the user specified parameters. a, b = (trunc_min - mu) / float(sigma), (trunc_max - mu) / float(sigma) sample = scipy.stats.truncnorm.rvs(a, b, mu, sigma, random_state=random_state) # scipy 1.5.1 returns an array while scipy 1.4.0 returns a scalar. # To maintain backwards compat we have to cast the sample to an array # then back to a scalar. return np.array(sample).item() def max_polyphony(ann): ''' Given an annotation of sound events, compute the maximum polyphony, i.e. the maximum number of simultaneous events at any given point in time. Only foreground events are taken into consideration for computing the polyphony. Parameters ---------- ann : JAMS.Annotation Returns ------- polyphony : int Maximum number of simultaneous events at any point in the annotation. ''' # If there are no foreground events the polyphony is 0 roles = [obs.value['role'] for obs in ann.data] if 'foreground' not in roles: return 0 else: # Keep only foreground events int_time, int_val = ann.to_interval_values() int_time_clean = [] for t, v in zip(int_time, int_val): if v['role'] == 'foreground': int_time_clean.append(t) int_time_clean = np.asarray(int_time_clean) # Sort and reshape arrivals = np.sort(int_time_clean[:, 0]).reshape(-1, 1) departures = np.sort(int_time_clean[:, 1]).reshape(-1, 1) # Onsets are +1, offsets are -1 arrivals = np.concatenate( (arrivals, np.ones(arrivals.shape)), axis=1) departures = np.concatenate( (departures, -np.ones(departures.shape)), axis=1) # Merge arrivals and departures and sort entry_log = np.concatenate((arrivals, departures), axis=0) entry_log_sorted = entry_log[entry_log[:, 0].argsort()] # Get maximum number of simultaneously occurring events polyphony = np.max(np.cumsum(entry_log_sorted[:, 1])) return int(polyphony) def polyphony_gini(ann, hop_size=0.01): ''' Compute the gini coefficient of the annotation's polyphony time series. Useful as an estimate of the polyphony "flatness" or entropy. The coefficient is in the range [0,1] and roughly inverse to entropy: a distribution that's close to uniform will have a low gini coefficient (high entropy), vice versa. https://en.wikipedia.org/wiki/Gini_coefficient Parameters ---------- ann : jams.Annotation Annotation for which to compute the normalized polyphony entropy. Must be of the scaper namespace. hop_size : float The hop size for sampling the polyphony time series. Returns ------- polyphony_gini: float Gini coefficient computed from the annotation's polyphony time series. Raises ------ ScaperError If the annotation does not have a duration value or if its namespace is not scaper. ''' if not ann.duration: raise ScaperError('Annotation does not have a duration value set.') if ann.namespace != 'scaper': raise ScaperError( 'Annotation namespace must be scaper, found {:s}.'.format( ann.namespace)) # If there are no foreground events the gini coefficient is 0 roles = [obs.value['role'] for obs in ann.data] if 'foreground' not in roles: return 0 # Sample the polyphony using the specified hop size n_samples = int(np.floor(ann.duration / float(hop_size)) + 1) times = np.linspace(0, (n_samples-1) * hop_size, n_samples) values = np.zeros_like(times) # for idx in ann.data.index: for obs in ann.data: # if ann.data.loc[idx, 'value']['role'] == 'foreground': if obs.value['role'] == 'foreground': start_time = obs.time end_time = start_time + obs.duration start_idx = np.argmin(np.abs(times - start_time)) end_idx = np.argmin(np.abs(times - end_time)) - 1 values[start_idx:end_idx + 1] += 1 values = values[:-1] # DEBUG # vstring = ('{:d} ' * len(values)).format(*tuple([int(v) for v in values])) # print(vstring) # print(' ') # Compute gini as per: # http://www.statsdirect.com/help/default.htm#nonparametric_methods/gini.htm values += 1e-6 # all values must be positive values = np.sort(values) # sort values n = len(values) i = np.arange(n) + 1 gini = np.sum((2*i - n - 1) * values) / (n * np.sum(values)) return (1 - gini) def is_real_number(num): ''' Check if a value is a real scalar by aggregating several numpy checks. Parameters ---------- num : any type The parameter to check Returns ------ check : bool True if ```num``` is a real scalar, False otherwise. ''' if (not np.isreal(num) or not np.isrealobj(num) or not np.isscalar(num)): return False else: return True def is_real_array(array): ''' Check if a value is a list or array of real scalars by aggregating several numpy checks. Parameters ---------- array: any type The parameter to check Returns ------ check : bool True if ```array``` is a list or array of a real scalars, False otherwise. ''' if not (type(array) is list or type(array) is np.ndarray): return False else: if (not np.all([np.isreal(x) for x in array]) or not np.isrealobj(array) or not np.asarray(list(map(np.isscalar, array))).all()): return False else: return True
15,371
28.964912
100
py
scaper
scaper-master/scaper/__init__.py
#!/usr/bin/env python """Top-level module for scaper""" from .core import Scaper from .core import generate_from_jams from .core import trim from .version import version as __version__
186
22.375
43
py
rpg_svo
rpg_svo-master/svo_analysis/setup.py
#!/usr/bin/env python from distutils.core import setup from catkin_pkg.python_setup import generate_distutils_setup d = generate_distutils_setup( packages=['svo_analysis'], package_dir={'': 'src'}, install_requires=['rospkg', 'yaml'], ) setup(**d)
266
21.25
60
py
rpg_svo
rpg_svo-master/svo_analysis/src/svo_analysis/filter_groundtruth_smooth.py
#!/usr/bin/python import numpy as np import matplotlib.pyplot as plt import transformations from scipy import signal save = True data_filename = '/home/cforster/Datasets/SlamBenchmark/asl_vicon_d2/groundtruth.txt' filtered_data_filename = '/home/cforster/Datasets/SlamBenchmark/asl_vicon_d2/groundtruth_filtered.txt' file = open(data_filename) data = file.read() lines = data.replace(","," ").replace("\t"," ").split("\n") D = np.array([[v.strip() for v in line.split(" ") if v.strip()!=""] for line in lines if len(line)>0 and line[0]!="#"], dtype=np.float64) n = np.shape(D)[0] rpy = np.empty([n,3]) for i in range(n): quat = D[i,4:8] rpy[i,:] = transformations.euler_from_quaternion(quat, axes='sxyz') # filter rpy f_sensor = 200.0; # sampling frequency in hz f_cut = 15.0; # cutoff frequency in hz b,a = signal.butter(5,f_cut/(f_sensor/2)); print a print b rpy_filt = np.empty([n,3]) rpy_filt[:,0] = signal.filtfilt(b, a, rpy[:,0]) rpy_filt[:,1] = signal.filtfilt(b, a, rpy[:,1]) rpy_filt[:,2] = signal.filtfilt(b, a, rpy[:,2]) fig = plt.figure() ax = fig.add_subplot(111, title='orientation filtered') ax.plot(rpy[:,0], 'r-') ax.plot(rpy[:,1], 'g-') ax.plot(rpy[:,2], 'b-') ax.plot(rpy_filt[:,0], 'k-', linewidth=2) ax.plot(rpy_filt[:,1], 'k-', linewidth=2) ax.plot(rpy_filt[:,2], 'k-', linewidth=2) fig = plt.figure() ax = fig.add_subplot(111, title='position') ax.plot(D[:,1], 'r') ax.plot(D[:,2], 'g') ax.plot(D[:,3], 'b') fig = plt.figure() ax = fig.add_subplot(111, title='trajectory from top') ax.plot(D[:,1], D[:,2]) if save: f = open(filtered_data_filename,'w') for i in range(np.shape(D)[0]): quat = transformations.quaternion_from_euler(rpy_filt[i,0], rpy_filt[i,1], rpy_filt[i,2], axes='sxyz') f.write('%.7f %.5f %.5f %.5f %.5f %.5f %.5f %.5f\n' % (D[i,0], D[i,1], D[i,2], D[i,3], quat[0], quat[1], quat[2], quat[3])) f.close()
1,875
29.754098
137
py
rpg_svo
rpg_svo-master/svo_analysis/src/svo_analysis/analyse_logs.py
#!/usr/bin/python import os import yaml import numpy as np import matplotlib.pyplot as plt def analyse_logs(D, trace_dir): # identify measurements which result from normal frames and which from keyframes is_kf = np.argwhere( (D['dropout'] == 1) & (D['repr_n_mps'] >= 0)) is_frame = np.argwhere(D['repr_n_mps'] >= 0) is_nokf = np.argwhere( (D['dropout'] == 0) & (D['repr_n_mps'] >= 0)) # set initial time to zero D['timestamp'] = D['timestamp'] - D['timestamp'][0] # ---------------------------------------------------------------------------- # plot number of reprojected points mean_n_reproj_points = np.mean(D['repr_n_mps'][is_frame]); mean_n_reproj_matches = np.mean(D['repr_n_new_references'][is_frame]); mean_n_edges_final = np.mean(D['sfba_n_edges_final'][is_frame]); fig = plt.figure(figsize=(8,3)) ax = fig.add_subplot(111, xlabel='time [s]') ax.plot(D['timestamp'][is_frame], D['repr_n_mps'][is_frame], 'r-', label='Reprojected Points, avg = %.2f'%mean_n_reproj_points) ax.plot(D['timestamp'][is_frame], D['repr_n_new_references'][is_frame], 'b-', label='Feature Matches, avg = %.2f'%mean_n_reproj_matches) ax.plot(D['timestamp'][is_frame], D['sfba_n_edges_final'][is_frame], 'g-', label='Points after Optimization, avg = %.2f'%mean_n_edges_final) ax.set_ylim(bottom=0) ax.legend(loc='lower right') fig.tight_layout() fig.savefig(os.path.join(trace_dir,'num_reprojected.pdf'), bbox_inches="tight") # ---------------------------------------------------------------------------- # plot median error before and after pose-optimzation and bundle adjustment init_error_avg = np.mean(D['sfba_error_init'][is_frame]) opt1_avg = np.mean(D['sfba_error_final'][is_frame]) fig = plt.figure(figsize=(8,2)) ax = fig.add_subplot(111, xlabel='time [s]', ylabel='error [px]') ax.plot(D['timestamp'][is_frame], D['sfba_error_init'][is_frame], 'r-', label='Initial error') ax.plot(D['timestamp'][is_frame], D['sfba_error_final'][is_frame], 'b-', label='Final error') ax.legend(ncol=2) fig.tight_layout() fig.savefig(os.path.join(trace_dir,'reprojection_error.pdf'), bbox_inches="tight") print 'average reprojection error improvement: ' + str(init_error_avg - opt1_avg) # ---------------------------------------------------------------------------- # plot number of candidate points fig = plt.figure(figsize=(8,3)) ax = fig.add_subplot(111, xlabel='time [s]') ax.plot(D['timestamp'][is_frame], D['n_candidates'][is_frame], 'r-', label='Candidate Points') fig.tight_layout() fig.savefig(os.path.join(trace_dir,'candidate_points.pdf'), bbox_inches="tight") # ---------------------------------------------------------------------------- # plot number of candidate points fig = plt.figure(figsize=(8,2)) ax = fig.add_subplot(111, xlabel='time [s]', ylabel='px') ax.plot(D['timestamp'][is_frame], D['sfba_thresh'][is_frame], 'r-', label='Threshold') fig.tight_layout() fig.savefig(os.path.join(trace_dir,'optimization_thresh.pdf'), bbox_inches="tight") # ---------------------------------------------------------------------------- # write other statistics to file stat = {'num_frames': len(is_frame), 'num_kfs': len(is_kf), 'reproj_error_avg_improvement': float(init_error_avg - opt1_avg)} with open(os.path.join(trace_dir,'dataset_stats.yaml'),'w') as outfile: outfile.write(yaml.dump(stat, default_flow_style=False))
3,497
45.64
96
py
rpg_svo
rpg_svo-master/svo_analysis/src/svo_analysis/analyse_timing.py
#!/usr/bin/python import os import numpy as np import matplotlib.pyplot as plt def analyse_timing(D, trace_dir): # identify measurements which result from normal frames and which from keyframes is_frame = np.argwhere(D['repr_n_mps'] >= 0) n_frames = len(is_frame) # set initial time to zero D['timestamp'] = D['timestamp'] - D['timestamp'][0] # ---------------------------------------------------------------------------- # plot total time for frame processing avg_time = np.mean(D['tot_time'][is_frame])*1000; fig = plt.figure(figsize=(8, 3)) ax = fig.add_subplot(111, ylabel='processing time [ms]', xlabel='time [s]') ax.plot(D['timestamp'][is_frame], D['tot_time'][is_frame]*1000, 'g-', label='total time [ms]') ax.plot(D['timestamp'][is_frame], np.ones(n_frames)*avg_time, 'b--', label=str('%(time).1fms mean time' % {'time': avg_time})) ax.legend() fig.tight_layout() fig.savefig(os.path.join(trace_dir,'timing.pdf'), bbox_inches="tight") # ---------------------------------------------------------------------------- # plot boxplot fig = plt.figure(figsize=(6,2)) ax = fig.add_subplot(111, xlabel='Processing time [ms]') ax.boxplot([ D['tot_time'][is_frame]*1000, # D['t_local_ba'][is_kf]*1000, D['pose_optimizer'][is_frame]*1000 + D['point_optimizer'][is_frame]*1000, D['reproject'][is_frame]*1000, D['sparse_img_align'][is_frame]*1000, D['pyramid_creation'][is_frame]*1000 ], 0,'', vert=0) boxplot_labels = [ r'\textbf{Total Motion Estimation: %2.2fms}' % np.median(D['tot_time'][is_frame]*1000), # 'Local BA (KF only): %.2fms ' % np.median(D['local_ba'][is_kf]*1000), 'Refinement: %2.2fms' % np.median(D['pose_optimizer'][is_frame]*1000 + D['point_optimizer'][is_frame]*1000), 'Feature Alignment: %2.2fms' % np.median(D['reproject'][is_frame]*1000), 'Sparse Image Alignment: %2.2fms' % np.median(D['sparse_img_align'][is_frame]*1000), 'Pyramid Creation: %2.2fms' % np.median(D['pyramid_creation'][is_frame]*1000) ] ax.set_yticks(np.arange(len(boxplot_labels))+1) ax.set_yticklabels(boxplot_labels) fig.tight_layout() fig.savefig(os.path.join(trace_dir,'timing_boxplot.pdf'), bbox_inches="tight") # ---------------------------------------------------------------------------- # plot boxplot reprojection fig = plt.figure(figsize=(6,2)) ax = fig.add_subplot(111, xlabel='Processing time [ms]') ax.boxplot([ D['reproject'][is_frame]*1000, D['feature_align'][is_frame]*1000, D['reproject_candidates'][is_frame]*1000, D['reproject_kfs'][is_frame]*1000 ], 0, '', vert=0) boxplot_labels = [r'\textbf{Total Reprojection: %2.2fms}' % np.median(D['reproject'][is_frame]*1000), 'Feature Alignment: %2.2fms' % np.median(D['feature_align'][is_frame]*1000), 'Reproject Candidates: %2.2fms' % np.median(D['reproject_candidates'][is_frame]*1000), 'Reproject Keyframes: %2.2fms' % np.median(D['reproject_kfs'][is_frame]*1000) ] ax.set_yticks(np.arange(len(boxplot_labels))+1) ax.set_yticklabels(boxplot_labels) fig.tight_layout() fig.savefig(os.path.join(trace_dir,'timing_reprojection.pdf'), bbox_inches="tight")
3,476
50.132353
128
py