| --- |
| language: |
| - en |
| - zh |
| tags: |
| - translation |
| license: cc-by-4.0 |
| datasets: |
| - quickmt/quickmt-train.zh-en |
| model-index: |
| - name: quickmt-en-zh |
| results: |
| - task: |
| name: Translation eng-zho |
| type: translation |
| args: eng-zho |
| dataset: |
| name: flores101-devtest |
| type: flores_101 |
| args: eng_Latn zho_Hans devtest |
| metrics: |
| - name: CHRF |
| type: chrf |
| value: 58.10 |
| - name: COMET |
| type: comet |
| value: 58.10 |
| --- |
| |
| <a href="https://huggingface.co/spaces/quickmt/quickmt-gui"><img src="https://huggingface.co/datasets/huggingface/badges/resolve/main/open-in-hf-spaces-lg-dark.svg" alt="Open in Spaces"></a> |
|
|
| # `quickmt-en-zh` Neural Machine Translation Model |
|
|
| `quickmt-en-zh` is a reasonably fast and reasonably accurate neural machine translation model for translation from `en` into `zh`. |
|
|
|
|
| ## Model Information |
|
|
| * Trained using [`eole`](https://github.com/eole-nlp/eole) |
| * 200M parameter transformer 'big' with 8 encoder layers and 2 decoder layers |
| * Separate source and target Sentencepiece tokenizers |
| * Exported for fast inference to [CTranslate2](https://github.com/OpenNMT/CTranslate2) format |
| * Training data: https://huggingface.co/datasets/quickmt/quickmt-train.zh-en/tree/main |
|
|
| See the `eole` model configuration in this repository for further details. |
|
|
|
|
| ## Usage with `quickmt` |
|
|
| You must install the Nvidia cuda toolkit first, if you want to do GPU inference. |
|
|
| Next, install the `quickmt` [python library](github.com/quickmt/quickmt). |
|
|
| ```bash |
| git clone https://github.com/quickmt/quickmt.git |
| pip install ./quickmt/ |
| ``` |
|
|
| Finally, use the model in python: |
|
|
| ```python |
| from quickmt import Translator |
| from huggingface_hub import snapshot_download |
| |
| # Download Model (if not downloaded already) and return path to local model |
| # Device is either 'auto', 'cpu' or 'cuda' |
| t = Translator( |
| snapshot_download("quickmt/quickmt-en-zh", ignore_patterns="eole-model/*"), |
| device="cpu" |
| ) |
| |
| # Translate - set beam size to 5 for higher quality (but slower speed) |
| t(["The Boot Monument is an American Revolutionary War memorial located in Saratoga National Historical Park in the state of New York."], beam_size=1) |
| |
| # Get alternative translations by sampling |
| # You can pass any cTranslate2 `translate_batch` arguments |
| t(["The Boot Monument is an American Revolutionary War memorial located in Saratoga National Historical Park in the state of New York."], sampling_temperature=1.2, beam_size=1, sampling_topk=50, sampling_topp=0.9) |
| ``` |
|
|
| The model is in `ctranslate2` format, and the tokenizers are `sentencepiece`, so you can use `ctranslate2` directly instead of through `quickmt`. It is also possible to get this model to work with e.g. [LibreTranslate](https://libretranslate.com/) which also uses `ctranslate2` and `sentencepiece`. |
|
|
|
|
| ## Metrics |
|
|
| `chrf2` is calculated with [sacrebleu](https://github.com/mjpost/sacrebleu) on the [Flores200 `devtest` test set](https://huggingface.co/datasets/facebook/flores) ("eng_Latn"->"zho_Hans"). `comet22` with the [`comet`](https://github.com/Unbabel/COMET) library and the [default model](https://huggingface.co/Unbabel/wmt22-comet-da). "Time (s)" is the time in seconds to translate (using `ctranslate2`) the flores-devtest dataset (1012 sentences) on an RTX 4070s GPU with batch size 32. |
|
|
| | Model | chrf2 | comet22 | Time (s) | |
| | -------------------------------- | ----- | ------- | -------- | |
| | quickmt/quickmt-en-zh | 35.22 | 85.39 | 0.96 | |
| | Helsinki-NLP/opus-mt-en-zh | 29.20 | 82.36 | 3.41 | |
| | facebook/m2m100_418M | 25.86 | 73.76 | 16.71 | |
| | facebook/m2m100_1.2B | 28.94 | 78.38 | 31.09 | |
| | facebook/nllb-200-distilled-600M | 24.52 | 78.41 | 19.01 | |
| | facebook/nllb-200-distilled-1.3B | 26.79 | 79.87 | 32.03 | |
|
|
| `quickmt-en-zh` is the fastest *and* highest quality. |
|
|
