# Fast-Inference with Ctranslate2

Speedup inference by 2x-8x using int8 inference in C++

quantized version of Helsinki-NLP/opus-mt-en-es

pip install hf-hub-ctranslate2>=1.0.0 ctranslate2>=3.13.0

Converted using

ct2-transformers-converter --model Helsinki-NLP/opus-mt-en-es --output_dir /home/michael/tmp-ct2fast-opus-mt-en-es --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 for device="cuda"
• compute_type=int8 for device="cpu"

from hf_hub_ctranslate2 import TranslatorCT2fromHfHub, GeneratorCT2fromHfHub
from transformers import AutoTokenizer

model_name = "michaelfeil/ct2fast-opus-mt-en-es"
# 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-en-es")
)
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

eng-spa

• source group: English

• target group: Spanish

• OPUS readme: eng-spa

• model: transformer

• source language(s): eng

• target language(s): spa

• model: transformer

• pre-processing: normalization + SentencePiece (spm32k,spm32k)

• download original weights: opus-2020-08-18.zip

• test set translations: opus-2020-08-18.test.txt

• test set scores: opus-2020-08-18.eval.txt

Benchmarks

System Info:

• hf_name: eng-spa

• source_languages: eng

• target_languages: spa

• opus_readme_url: https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/eng-spa/README.md

• original_repo: Tatoeba-Challenge

• tags:

• ctranslate2 ['translation']

• languages: ['en', 'es']

• src_constituents: {'eng'}

• tgt_constituents: {'spa'}

• src_multilingual: False

• tgt_multilingual: False

• prepro: normalization + SentencePiece (spm32k,spm32k)

• url_model: https://object.pouta.csc.fi/Tatoeba-MT-models/eng-spa/opus-2020-08-18.zip

• url_test_set: https://object.pouta.csc.fi/Tatoeba-MT-models/eng-spa/opus-2020-08-18.test.txt

• src_alpha3: eng

• tgt_alpha3: spa

• short_pair: en-es

• chrF2_score: 0.721

• bleu: 54.9

• brevity_penalty: 0.978

• ref_len: 77311.0

• src_name: English

• tgt_name: Spanish

• train_date: 2020-08-18 00:00:00

• src_alpha2: en

• tgt_alpha2: es

• prefer_old: False

• long_pair: eng-spa

• helsinki_git_sha: d2f0910c89026c34a44e331e785dec1e0faa7b82

• transformers_git_sha: f7af09b4524b784d67ae8526f0e2fcc6f5ed0de9

• port_machine: brutasse

• port_time: 2020-08-24-18:20