Instructions to use hf-internal-testing/tiny-xlm-roberta with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use hf-internal-testing/tiny-xlm-roberta with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="hf-internal-testing/tiny-xlm-roberta")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("hf-internal-testing/tiny-xlm-roberta")
model = AutoModelForCausalLM.from_pretrained("hf-internal-testing/tiny-xlm-roberta")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use hf-internal-testing/tiny-xlm-roberta with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "hf-internal-testing/tiny-xlm-roberta"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "hf-internal-testing/tiny-xlm-roberta",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/hf-internal-testing/tiny-xlm-roberta

SGLang

How to use hf-internal-testing/tiny-xlm-roberta with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "hf-internal-testing/tiny-xlm-roberta" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "hf-internal-testing/tiny-xlm-roberta",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "hf-internal-testing/tiny-xlm-roberta" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "hf-internal-testing/tiny-xlm-roberta",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use hf-internal-testing/tiny-xlm-roberta with Docker Model Runner:
```
docker model run hf.co/hf-internal-testing/tiny-xlm-roberta
```

stas commited on Jul 8, 2021

Commit

1e39576

1 Parent(s): d361e26

updates

Browse files

Files changed (2) hide show

make-tiny-xlm-roberta.py +2 -6
pytorch_model.bin +1 -1

make-tiny-xlm-roberta.py CHANGED Viewed

@@ -60,17 +60,11 @@
 # git commit -m "new tiny model"
 # git push
-from pathlib import Path
-import json
-import tempfile
 import sys
 import os
 from transformers import XLMRobertaTokenizer, XLMRobertaTokenizerFast, XLMRobertaConfig, XLMRobertaForCausalLM
-# workaround for fast tokenizer protobuffer issue, and it's much faster too!
-os.environ["PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION"] = "python"
 mname_orig = "xlm-roberta-base"
 mname_tiny = "tiny-xlm-roberta"
@@ -81,6 +75,8 @@ vocab_keep_items = 5000
 tmp_dir = f"/tmp/{mname_tiny}"
 vocab_orig_path = f"{tmp_dir}/sentencepiece.bpe.model"
 vocab_short_path = f"{tmp_dir}/spiece-short.model"
 if 1: # set to 0 to skip this after running once to speed things up during tune up
     # HACK: need the sentencepiece source to get sentencepiece_model_pb2, as it doesn't get installed
     sys.path.append("../sentencepiece/python/src/sentencepiece")

 # git commit -m "new tiny model"
 # git push
 import sys
 import os
 from transformers import XLMRobertaTokenizer, XLMRobertaTokenizerFast, XLMRobertaConfig, XLMRobertaForCausalLM
 mname_orig = "xlm-roberta-base"
 mname_tiny = "tiny-xlm-roberta"
 tmp_dir = f"/tmp/{mname_tiny}"
 vocab_orig_path = f"{tmp_dir}/sentencepiece.bpe.model"
 vocab_short_path = f"{tmp_dir}/spiece-short.model"
+# workaround for fast tokenizer protobuf issue, and it's much faster too!
+os.environ["PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION"] = "python"
 if 1: # set to 0 to skip this after running once to speed things up during tune up
     # HACK: need the sentencepiece source to get sentencepiece_model_pb2, as it doesn't get installed
     sys.path.append("../sentencepiece/python/src/sentencepiece")

pytorch_model.bin CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b9c6cf79904c41a0ee0dd56366d40b9d2762235495edc6d04693ca3f41c50052
 size 4334436

 version https://git-lfs.github.com/spec/v1
+oid sha256:eb0bccafb4bee811f2138956ea9e94596e1bfdfc868b5364d7b678fac4b2d559
 size 4334436