Note

NMT research done during time in SIL International. Final working copy has been integrated to SIL's repo. Work here is no longer mantained/updated.

Setup

1. Create a Hugging Face account at https://huggingface.co/
2. Add a .env file at root of repository and add SIL_NLP_DATA_PATH="S:"
3. Create a new user access token at https://huggingface.co/settings/tokens. For the 'role' option chose 'write', to allow you to upload a HF model to the HF hub.
4. Retrieve your AWS s3 credentials for 'key' and 'secret' from your clearml.conf file.
5. On your ClearML account, go to Settings>Workspace>Configuration Vault. Add your HF access token and AWS credentials. This gives ClearML access to HF and the s3 bucket.

huggingface { token: "" }

aws { s3 { region: "" key: "" secret: "" } }

5. In your IDE environment, set your Python Intepreter to python 3.8.10.
6. Set your project name and task name in the codebase.
7. Launch your experiment in the CLI. Prior to launching the experiment, you may need to download relevant Python packages.
8. On the ClearML GUI, clone the experiment/reset it.
9. In ClearML GUI, Experiments>Execution>Container make the following changes. Experiment must be a "Draft" to make edits.

• Set Container>Image to 'python:3:8.10'
• In Container> Setup Shell Script add:

apt-get install git-lfs git lfs install apt-get install python3-distutils

10. Enqueue the experiment

Notes

1. HF access tokens : https://huggingface.co/docs/hub/security-tokens
2. Download Python 3.8.10 at https://www.python.org/downloads/release/python-3810/
3. Package dependencies for ClearML are specified in requirements.txt.
4. Launching an experiment in the CLI is to register the experiment on ClearML. Ctrl+C when appropriate in CLI and enqueue on ClearML GUI.
5. Under Configurations in the ClearML GUI, expect to see Hyperparameters being updated.

Translation

This directory contains examples for finetuning and evaluating transformers on translation tasks. Please tag @patil-suraj with any issues/unexpected behaviors, or send a PR! For deprecated bertabs instructions, see bertabs/README.md. For the old finetune_trainer.py and related utils, see examples/legacy/seq2seq.

Supported Architectures

• BartForConditionalGeneration
• FSMTForConditionalGeneration (translation only)
• MBartForConditionalGeneration
• MarianMTModel
• PegasusForConditionalGeneration
• T5ForConditionalGeneration
• MT5ForConditionalGeneration

run_translation.py is a lightweight examples of how to download and preprocess a dataset from the 🤗 Datasets library or use your own files (jsonlines or csv), then fine-tune one of the architectures above on it.

For custom datasets in jsonlines format please see: https://huggingface.co/docs/datasets/loading_datasets.html#json-files and you also will find examples of these below.

With Trainer

Here is an example of a translation fine-tuning with a MarianMT model:

python examples/pytorch/translation/run_translation.py \
    --model_name_or_path Helsinki-NLP/opus-mt-en-ro \
    --do_train \
    --do_eval \
    --source_lang en \
    --target_lang ro \
    --dataset_name wmt16 \
    --dataset_config_name ro-en \
    --output_dir /tmp/tst-translation \
    --per_device_train_batch_size=4 \
    --per_device_eval_batch_size=4 \
    --overwrite_output_dir \
    --predict_with_generate

MBart and some T5 models require special handling.

T5 models t5-small, t5-base, t5-large, t5-3b and t5-11b must use an additional argument: --source_prefix "translate {source_lang} to {target_lang}". For example:

python examples/pytorch/translation/run_translation.py \
    --model_name_or_path t5-small \
    --do_train \
    --do_eval \
    --source_lang en \
    --target_lang ro \
    --source_prefix "translate English to Romanian: " \
    --dataset_name wmt16 \
    --dataset_config_name ro-en \
    --output_dir /tmp/tst-translation \
    --per_device_train_batch_size=4 \
    --per_device_eval_batch_size=4 \
    --overwrite_output_dir \
    --predict_with_generate

If you get a terrible BLEU score, make sure that you didn't forget to use the --source_prefix argument.

For the aforementioned group of T5 models it's important to remember that if you switch to a different language pair, make sure to adjust the source and target values in all 3 language-specific command line argument: --source_lang, --target_lang and --source_prefix.

MBart models require a different format for --source_lang and --target_lang values, e.g. instead of en it expects en_XX, for ro it expects ro_RO. The full MBart specification for language codes can be found here. For example:

python examples/pytorch/translation/run_translation.py \
    --model_name_or_path facebook/mbart-large-en-ro  \
    --do_train \
    --do_eval \
    --dataset_name wmt16 \
    --dataset_config_name ro-en \
    --source_lang en_XX \
    --target_lang ro_RO \
    --output_dir /tmp/tst-translation \
    --per_device_train_batch_size=4 \
    --per_device_eval_batch_size=4 \
    --overwrite_output_dir \
    --predict_with_generate

And here is how you would use the translation finetuning on your own files, after adjusting the values for the arguments --train_file, --validation_file to match your setup:

python examples/pytorch/translation/run_translation.py \
    --model_name_or_path t5-small \
    --do_train \
    --do_eval \
    --source_lang en \
    --target_lang ro \
    --source_prefix "translate English to Romanian: " \
    --dataset_name wmt16 \
    --dataset_config_name ro-en \
    --train_file path_to_jsonlines_file \
    --validation_file path_to_jsonlines_file \
    --output_dir /tmp/tst-translation \
    --per_device_train_batch_size=4 \
    --per_device_eval_batch_size=4 \
    --overwrite_output_dir \
    --predict_with_generate

The task of translation supports only custom JSONLINES files, with each line being a dictionary with a key "translation" and its value another dictionary whose keys is the language pair. For example:

{ "translation": { "en": "Others have dismissed him as a joke.", "ro": "Alții l-au numit o glumă." } }
{ "translation": { "en": "And some are holding out for an implosion.", "ro": "Iar alții așteaptă implozia." } }

Here the languages are Romanian (ro) and English (en).

If you want to use a pre-processed dataset that leads to high BLEU scores, but for the en-de language pair, you can use --dataset_name stas/wmt14-en-de-pre-processed, as following:

python examples/pytorch/translation/run_translation.py \
    --model_name_or_path t5-small \
    --do_train \
    --do_eval \
    --source_lang en \
    --target_lang de \
    --source_prefix "translate English to German: " \
    --dataset_name stas/wmt14-en-de-pre-processed \
    --output_dir /tmp/tst-translation \
    --per_device_train_batch_size=4 \
    --per_device_eval_batch_size=4 \
    --overwrite_output_dir \
    --predict_with_generate

With Accelerate

Based on the script run_translation_no_trainer.py.

Like run_translation.py, this script allows you to fine-tune any of the models supported on a translation task, the main difference is that this script exposes the bare training loop, to allow you to quickly experiment and add any customization you would like.

It offers less options than the script with Trainer (for instance you can easily change the options for the optimizer or the dataloaders directly in the script) but still run in a distributed setup, on TPU and supports mixed precision by the mean of the 🤗 Accelerate library. You can use the script normally after installing it:

pip install accelerate

then

python run_translation_no_trainer.py \
    --model_name_or_path Helsinki-NLP/opus-mt-en-ro \
    --source_lang en \
    --target_lang ro \
    --dataset_name wmt16 \
    --dataset_config_name ro-en \
    --output_dir ~/tmp/tst-translation

You can then use your usual launchers to run in it in a distributed environment, but the easiest way is to run

accelerate config

and reply to the questions asked. Then

accelerate test

that will check everything is ready for training. Finally, you can launch training with

accelerate launch run_translation_no_trainer.py \
    --model_name_or_path Helsinki-NLP/opus-mt-en-ro \
    --source_lang en \
    --target_lang ro \
    --dataset_name wmt16 \
    --dataset_config_name ro-en \
    --output_dir ~/tmp/tst-translation

This command is the same and will work for:

• a CPU-only setup
• a setup with one GPU
• a distributed training with several GPUs (single or multi node)
• a training on TPUs

Note that this library is in alpha release so your feedback is more than welcome if you encounter any problem using it.