Skip to content

User Guide

Quick links:

💻 GitHub Repository
📚 Documentation
🐍 PyPi project
🧪 Colab Demo / Kaggle Demo

Quickstart

Install the library with pip:

pip install dl-translate

To translate some text:

import dl_translate as dlt

mt = dlt.TranslationModel()  # Slow when you load it for the first time

text_hi = "संयुक्त राष्ट्र के प्रमुख का कहना है कि सीरिया में कोई सैन्य समाधान नहीं है"
mt.translate(text_hi, source=dlt.lang.HINDI, target=dlt.lang.ENGLISH)

Above, you can see that dlt.lang contains variables representing each of the 50 available languages with auto-complete support. Alternatively, you can specify the language (e.g. "Arabic") or the language code (e.g. "fr" for French):

text_ar = "الأمين العام للأمم المتحدة يقول إنه لا يوجد حل عسكري في سوريا."
mt.translate(text_ar, source="Arabic", target="fr")

If you want to verify whether a language is available, you can check it:

print(mt.available_languages())  # All languages that you can use
print(mt.available_codes())  # Code corresponding to each language accepted
print(mt.get_lang_code_map())  # Dictionary of lang -> code

Usage

Selecting a device

When you load the model, you can specify the device using the device argument. By default, the value will be device="auto", which means it will use a GPU if possible. You can also explicitly set device="cpu" or device="gpu", or some other strings accepted by torch.device(). In general, it is recommend to use a GPU if you want a reasonable processing time.

mt = dlt.TranslationModel(device="auto")  # Automatically select device
mt = dlt.TranslationModel(device="cpu")  # Force you to use a CPU
mt = dlt.TranslationModel(device="gpu")  # Force you to use a GPU
mt = dlt.TranslationModel(device="cuda:2")  # Use the 3rd GPU available

Choosing a different model

By default, the m2m100 model will be used. However, there are a few options:

  • mBART-50 Large: Allows translations across 50 languages.
  • m2m100: Allows translations across 100 languages.
  • nllb-200 (New in v0.3): Allows translations across 200 languages, and is faster than m2m100 (On RTX A6000, we can see speed up of 3x).

Here's an example:

# The default approval
mt = dlt.TranslationModel("m2m100")  # Shorthand
mt = dlt.TranslationModel("facebook/m2m100_418M")  # Huggingface repo

# If you want to use mBART-50 Large
mt = dlt.TranslationModel("mbart50")
mt = dlt.TranslationModel("facebook/mbart-large-50-many-to-many-mmt")

# Or NLLB-200 (faster and has 200 languages)
mt = dlt.TranslationModel("nllb200")
mt = dlt.TranslationModel("facebook/nllb-200-distilled-600M")

Note that the language code will change depending on the model family. To find out the correct language codes, please read the doc page on available languages or run mt.available_codes().

By default, dlt.TranslationModel will download the model from the huggingface repo for mbart50, m2m100, or nllb200 and cache it. It's possible to load the model from a path or a model with a similar format, but you will need to specify the model_family:

mt = dlt.TranslationModel("/path/to/model/directory/", model_family="mbart50")
mt = dlt.TranslationModel("facebook/m2m100_1.2B", model_family="m2m100")
mt = dlt.TranslationModel("facebook/nllb-200-distilled-600M", model_family="nllb200")

Notes: * Make sure your tokenizer is also stored in the same directory if you load from a file. * The available languages will change if you select a different model, so you will not be able to leverage dlt.lang or dlt.utils.

Breaking down into sentences

It is not recommended to use extremely long texts as it takes more time to process. Instead, you can try to break them down into sentences. Multiple solutions exists for that, including doing it manually and using the nltk library.

A quick approach would be to split them by period. However, you have to ensure that there are no periods used for abbreviations (such as Mr. or Dr.). For example, it will work in the following case:

text = "Mr Smith went to his favorite cafe. There, he met his friend Dr Doe."
sents = text.split(".")
".".join(mt.translate(sents, source=dlt.lang.ENGLISH, target=dlt.lang.FRENCH))

For more complex cases (e.g. where you use periods for abbreviations), you can use nltk. First install the library with pip install nltk, then run:

import nltk

nltk.download("punkt")

text = "Mr. Smith went to his favorite cafe. There, he met his friend Dr. Doe."
sents = nltk.tokenize.sent_tokenize(text, "english")  # don't use dlt.lang.ENGLISH
" ".join(mt.translate(sents, source=dlt.lang.ENGLISH, target=dlt.lang.FRENCH))

Batch size and verbosity when using translate

It's possible to set a batch size (i.e. the number of elements processed at once) for mt.translate and whether you want to see the progress bar or not:

...
mt = dlt.TranslationModel()
mt.translate(text, source, target, batch_size=32, verbose=True)

If you set batch_size=None, it will compute the entire text at once rather than splitting into "chunks". We recommend lowering batch_size if you do not have a lot of RAM or VRAM and run into CUDA memory error. Set a higher value if you are using a high-end GPU and the VRAM is not fully utilized.

dlt.utils module

An alternative to mt.available_languages() is the dlt.utils module. You can use it to find out which languages and codes are available:

print(dlt.utils.available_languages('mbart50'))  # All languages that you can use
print(dlt.utils.available_codes('mbart50'))  # Code corresponding to each language accepted
print(dlt.utils.get_lang_code_map('mbart50'))  # Dictionary of lang -> code
print(dlt.utils.available_languages('m2m100'))  # write the name of the model family

At the moment, the following models are accepted: - "mbart50" - "m2m100" - "nllb200"

Offline usage

Unlike the Google translate or MSFT Translator APIs, this library can be fully used offline. However, you will need to first download the packages and models, and move them to your offline environment to be installed and loaded inside a venv.

First, run in your terminal:

mkdir dlt
cd dlt
mkdir libraries
pip download -d libraries/ dl-translate

Once all the required packages are downloaded, you will need to use huggingface hub to download the files. Install it with pip install huggingface-hub. Then, run inside Python:

import shutil
import huggingface_hub as hub

dirname = hub.snapshot_download("facebook/m2m100_418M")
shutil.copytree(dirname, "cached_model_m2m100")  # Copy to a permanent folder

Now, move everything in the dlt directory to your offline environment. Create a virtual environment and run the following in terminal:

pip install --no-index --find-links libraries/ dl-translate

Now, run inside Python:

import dl_translate as dlt

mt = dlt.TranslationModel("cached_model_m2m100", model_family="m2m100")

Advanced

The following section assumes you have knowledge of PyTorch and Huggingface Transformers.

Saving and loading

If you wish to accelerate the loading time the translation model, you can use save_obj. Later you can reload it with load_obj by specifying the same directory that you are using to save.

mt = dlt.TranslationModel()
# ...
mt.save_obj('saved_model')
# ...
mt = dlt.TranslationModel.load_obj('saved_model')

Warning: Only use this if you are certain the torch module saved in saved_model/weights.pt can be correctly loaded. Indeed, it is possible that the huggingface, torch or some other dependencies change between when you called save_obj and load_obj, and that might break your code. Thus, it is recommend to only run load_obj in the same environment/session as save_obj. Note this method might be deprecated in the future once there's no speed benefit in loading this way.

Interacting with underlying model and tokenizer

When initializing model, you can pass in arguments for the underlying BART model and tokenizer (which will respectively be passed to ModelForConditionalGeneration.from_pretrained and TokenizerFast.from_pretrained):

mt = dlt.TranslationModel(
    model_options=dict(
        state_dict=...,
        cache_dir=...,
        ...
    ),
    tokenizer_options=dict(
        tokenizer_file=...,
        eos_token=...,
        ...
    )
)

You can also access the underlying transformers model and tokenizer:

transformers_model = mt.get_transformers_model()
tokenizer = mt.get_tokenizer()

For more information about the models themselves, please read the docs on mBART and m2m100.

Keyword arguments for the generate() method of the underlying model

When running mt.translate, you can also give a generation_options dictionary that is passed as keyword arguments to the underlying mt.get_transformers_model().generate() method:

mt.translate(
    text,
    source=dlt.lang.GERMAN,
    target=dlt.lang.SPANISH,
    generation_options=dict(num_beams=5, max_length=...)
)

Learn more in the huggingface docs.