Datasets:
jblitzar
/
github-python-corpus

Dataset card Data Studio Files
xet
 Community
text
stringlengths
1
93.6k
parser.add_argument('--silent', action='store_true',
help='Flag to supress training progress bar for each epoch')
parser.add_argument('--lr', type=float, default=0.0001)
parser.add_argument('--epochs', type=int, default=100)
parser.add_argument('--bsz', type=int, default=128)
parser.add_argument('--ckpt', type=str, required=True,
help='filepath at which to save best probing model (on dev set)')
parser.add_argument('--encoder-name', type=str, default='bert-base',
choices=['bert-base', 'bert-large', 'roberta-base', 'roberta-large'])
parser.add_argument('--kshot', type=int, default=-1,
help='if set to k (1+), will filter training data to only have up to k examples per sense')
parser.add_argument('--data-path', type=str, required=True,
help='Location of top-level directory for the Unified WSD Framework')
parser.add_argument('--eval', action='store_true',
help='Flag to set script to evaluate probe (rather than train)')
parser.add_argument('--split', type=str, default='semeval2007',
choices=['semeval2007', 'senseval2', 'senseval3', 'semeval2013', 'semeval2015', 'ALL', 'all-test'],
help='Which evaluation split on which to evaluate probe')
def wn_keys(data):
keys = []
for sent in data:
for form, lemma, pos, inst, _ in sent:
if inst != -1:
key = generate_key(lemma, pos)
keys.append(key)
return keys
def batchify(data, bsz=1):
print('Batching data with bsz={}...'.format(bsz))
batched_data = []
for i in range(0, len(data), bsz):
if i+bsz < len(data): d_arr = data[i:i+bsz]
else: d_arr = data[i:] #get remainder examples
batched_ids = torch.cat([ids for ids, _, _, _ in d_arr], dim=0)
batched_masks = torch.stack([mask for _, mask, _, _ in d_arr], dim=0)
batched_insts = [inst for _, _, inst, _ in d_arr]
batched_labels = torch.cat([label for _, _, _, label in d_arr], dim=0)
batched_data.append((batched_ids, batched_masks, batched_insts, batched_labels))
return batched_data
#takes in text data, tensorizes it for BERT, runs though BERT,
#filters out the context words (not labeled), and averages
#the representation(s) for words/phrases to be disambiguated
#output is tuples of (input tensor prepared for linear probing model,
#instance numbers (for dataset), tensor of label indexes)
def preprocess(tokenizer, context_model, text_data, label_space, label_map):
processed_examples = []
output_masks = []
instances = []
label_indexes = []
#tensorize data
for sent in tqdm(text_data):
sent_ids = [torch.tensor([tokenizer.encode(tokenizer.cls_token)])] #aka sos token, returns a list with single index
bert_mask = [-1]
for idx, (word, lemma, pos, inst, label) in enumerate(sent):
word_ids = torch.tensor([tokenizer.encode(word.lower())])
sent_ids.append(word_ids)
if inst != -1:
#masking for averaging of bert outputs
bert_mask.extend([idx]*word_ids.size(-1))
#tracking instance for sense-labeled word
instances.append(inst)
#adding label tensor for senes-labeled word
if label in label_space:
label_indexes.append(torch.tensor([label_space.index(label)]))
else:
label_indexes.append(torch.tensor([label_space.index('n/a')]))
#adding appropriate label space for sense-labeled word (we only use this for wsd task)
key = generate_key(lemma, pos)
if key in label_map:
l_space = label_map[key]
o_mask = torch.zeros(len(label_space))
for l in l_space: o_mask[l] = 1
output_masks.append(o_mask)
else:
output_masks.append(torch.ones(len(label_space))) #let this predict whatever -- should not use this (default to backoff for unseen forms)
else:
bert_mask.extend([-1]*word_ids.size(-1))
#add eos token
sent_ids.append(torch.tensor([tokenizer.encode(tokenizer.sep_token)])) #aka eos token
bert_mask.append(-1)
sent_ids = torch.cat(sent_ids, dim=-1)
#run inputs through frozen bert
sent_ids = sent_ids.cuda()
with torch.no_grad():
output = context_model(sent_ids)[0].squeeze().cpu()
#average outputs for subword units in same word/phrase, drop unlabeled words