utrobinmv's picture
feat update readme
5a49c01
metadata
language:
  - ru
  - zh
  - en
tags:
  - translation
  - text2text-generation
  - t5
license: apache-2.0
datasets:
  - ccmatrix
metrics:
  - sacrebleu
widget:
  - example_title: translate zh-ru
    text: |
      translate to ru: 开发的目的是为用户提供个人同步翻译。
  - example_title: translate ru-en
    text: >
      translate to en: Цель разработки — предоставить пользователям личного
      синхронного переводчика.
  - example_title: translate en-ru
    text: >
      translate to ru: The purpose of the development is to provide users with a
      personal synchronized interpreter.
  - example_title: translate en-zh
    text: >
      translate to zh: The purpose of the development is to provide users with a
      personal synchronized interpreter.
  - example_title: translate zh-en
    text: |
      translate to en: 开发的目的是为用户提供个人同步解释器。
  - example_title: translate ru-zh
    text: >
      translate to zh: Цель разработки — предоставить пользователям личного
      синхронного переводчика.
model-index:
  - name: utrobinmv/t5_translate_en_ru_zh_base_200
    results:
      - task:
          type: translation
          name: Translation en-ru
        dataset:
          name: ntrex_en-ru
          type: ntrex
          config: ntrex en-ru
          split: test
        metrics:
          - type: sacrebleu
            value: 28.575940911021487
            name: bleu
            verified: false
          - type: chrf
            value: 54.27996346886896
            name: chrf
            verified: false
          - type: ter
            value: 62.494863914873584
            name: ter
            verified: false
          - type: meteor
            value: 0.5174833677740809
            name: meteor
            verified: false
          - type: rouge
            value: 0.1908317951570274
            name: ROUGE-1
            verified: false
          - type: rouge
            value: 0.065555552204933
            name: ROUGE-2
            verified: false
          - type: rouge
            value: 0.1895542893295215
            name: ROUGE-L
            verified: false
          - type: rouge
            value: 0.1893813749889601
            name: ROUGE-LSUM
            verified: false
          - type: bertscore
            value: 0.8554933660030365
            name: bertscore_f1
            verified: false
          - type: bertscore
            value: 0.8578473615646363
            name: bertscore_precision
            verified: false
          - type: bertscore
            value: 0.8534188346862793
            name: bertscore_recall
            verified: false
        source:
          name: NTREX dataset Benchmark
          url: https://huggingface.co/spaces/utrobinmv/TREX_benchmark_en_ru_zh
  - name: utrobinmv/t5_translate_en_ru_zh_base_200
    results:
      - task:
          type: translation
          name: Translation ru-en
        dataset:
          name: ntrex_ru-en
          type: ntrex
          config: ntrex ru-en
          split: test
        metrics:
          - type: sacrebleu
            value: 28.575940911021487
            name: bleu
            verified: false
          - type: chrf
            value: 54.27996346886896
            name: chrf
            verified: false
          - type: ter
            value: 62.494863914873584
            name: ter
            verified: false
          - type: meteor
            value: 0.5174833677740809
            name: meteor
            verified: false
          - type: rouge
            value: 0.1908317951570274
            name: ROUGE-1
            verified: false
          - type: rouge
            value: 0.065555552204933
            name: ROUGE-2
            verified: false
          - type: rouge
            value: 0.1895542893295215
            name: ROUGE-L
            verified: false
          - type: rouge
            value: 0.1893813749889601
            name: ROUGE-LSUM
            verified: false
          - type: bertscore
            value: 0.8554933660030365
            name: bertscore_f1
            verified: false
          - type: bertscore
            value: 0.8578473615646363
            name: bertscore_precision
            verified: false
          - type: bertscore
            value: 0.8534188346862793
            name: bertscore_recall
            verified: false
        source:
          name: NTREX dataset Benchmark
          url: https://huggingface.co/spaces/utrobinmv/TREX_benchmark_en_ru_zh

T5 English, Russian and Chinese multilingual machine translation

This model represents a conventional T5 transformer in multitasking mode for translation into the required language, precisely configured for machine translation for pairs: ru-zh, zh-ru, en-zh, zh-en, en-ru, ru-en.

The model can perform direct translation between any pair of Russian, Chinese or English languages. For translation into the target language, the target language identifier is specified as a prefix 'translate to :'. In this case, the source language may not be specified, in addition, the source text may be multilingual.

Example translate Russian to Chinese

from transformers import T5ForConditionalGeneration, T5Tokenizer

model_name = 'utrobinmv/t5_translate_en_ru_zh_small_1024'
model = T5ForConditionalGeneration.from_pretrained(model_name)
tokenizer = T5Tokenizer.from_pretrained(model_name)

prefix = 'translate to zh: '
src_text = prefix + "Цель разработки — предоставить пользователям личного синхронного переводчика."

# translate Russian to Chinese
input_ids = tokenizer(src_text, return_tensors="pt")

generated_tokens = model.generate(**input_ids)

result = tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)
print(result)
#开发的目的是为用户提供个人同步翻译。

and Example translate Chinese to Russian

from transformers import T5ForConditionalGeneration, T5Tokenizer

model_name = 'utrobinmv/t5_translate_en_ru_zh_small_1024'
model = T5ForConditionalGeneration.from_pretrained(model_name)
tokenizer = T5Tokenizer.from_pretrained(model_name)

prefix = 'translate to ru: '
src_text = prefix + "开发的目的是为用户提供个人同步翻译。"

# translate Russian to Chinese
input_ids = tokenizer(src_text, return_tensors="pt")

generated_tokens = model.generate(**input_ids)

result = tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)
print(result)
#Цель разработки - предоставить пользователям персональный синхронный перевод.

Languages covered

Russian (ru_RU), Chinese (zh_CN), English (en_US)