# Fast-Inference with Ctranslate2
Speedup inference by 2x-8x using int8 inference in C++
quantized version of Helsinki-NLP/opus-mt-de-fr
pip install hf-hub-ctranslate2>=1.0.0 ctranslate2>=3.13.0
Converted using
ct2-transformers-converter --model Helsinki-NLP/opus-mt-de-fr --output_dir /home/michael/tmp-ct2fast-opus-mt-de-fr --force --copy_files README.md generation_config.json tokenizer_config.json vocab.json source.spm .gitattributes target.spm --quantization float16
Checkpoint compatible to ctranslate2 and hf-hub-ctranslate2
compute_type=int8_float16
fordevice="cuda"
compute_type=int8
fordevice="cpu"
from hf_hub_ctranslate2 import TranslatorCT2fromHfHub, GeneratorCT2fromHfHub
from transformers import AutoTokenizer
model_name = "michaelfeil/ct2fast-opus-mt-de-fr"
# use either TranslatorCT2fromHfHub or GeneratorCT2fromHfHub here, depending on model.
model = TranslatorCT2fromHfHub(
# load in int8 on CUDA
model_name_or_path=model_name,
device="cuda",
compute_type="int8_float16",
tokenizer=AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-de-fr")
)
outputs = model.generate(
text=["How do you call a fast Flan-ingo?", "User: How are you doing?"],
)
print(outputs)
Licence and other remarks:
This is just a quantized version. Licence conditions are intended to be idential to original huggingface repo.
Original description
opus-mt-de-fr
source languages: de
target languages: fr
OPUS readme: de-fr
dataset: opus
model: transformer-align
pre-processing: normalization + SentencePiece
download original weights: opus-2020-01-08.zip
test set translations: opus-2020-01-08.test.txt
test set scores: opus-2020-01-08.eval.txt
Benchmarks
testset | BLEU | chr-F |
---|---|---|
euelections_dev2019.transformer-align.de | 32.2 | 0.590 |
newssyscomb2009.de.fr | 26.8 | 0.553 |
news-test2008.de.fr | 26.4 | 0.548 |
newstest2009.de.fr | 25.6 | 0.539 |
newstest2010.de.fr | 29.1 | 0.572 |
newstest2011.de.fr | 26.9 | 0.551 |
newstest2012.de.fr | 27.7 | 0.554 |
newstest2013.de.fr | 29.5 | 0.560 |
newstest2019-defr.de.fr | 36.6 | 0.625 |
Tatoeba.de.fr | 49.2 | 0.664 |
- Downloads last month
- 14
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.