This repository contains pretraining pipeline of sequence-to-sequence language models.

We provide scripts based on the Fairseq library and PyTorch.

This codebase requires CUDA 11.3+, Python 3.8+, and PyTorch 1.10.2+. For best compatibility, you can run the following script in a clean Python 3.8 virtual environment:

python -m pip install torch==1.10.2+cu113 torchaudio==0.10.2+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html

Another dependency is Fairseq.

git clone [email protected]:facebookresearch/fairseq
cd fairseq
git checkout 11b2830d29aed8043e5011d64e14004347a08b50
python -m pip install -e .

Please refer to Fairseq.

The following pretraining script for pretraining T5-Base on the Wikibook dataset is tested on a node of 8 NVIDIA A100 40GB GPUs:

export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
fairseq-hydra-train -m --config-dir examples/t5/config/pretraining \
--config-name t5_base_8gpus \
common.user_dir=$(pwd)/efficent_large_lm_trainer \
task.data=/path/to/wikibook_data \
hydra.sweep.dir=/path/to/outputs

Pretraining on a single node will take ~136 hours. We recommend pretraining on 8 nodes. Assuming the NODE_RANK environment variable is set to i on the i-th node, here is the pretraining script on 8 nodes with 8 GPUs each:

export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
fairseq-hydra-train -m --config-dir examples/t5/config/pretraining \
--config-name t5_base_64gpus \
common.user_dir=$(pwd)/efficent_large_lm_trainer \
task.data=/path/to/wikibook_data \
distributed_training.distributed_world_size=64 \
distributed_training.distributed_rank=$((NODE_RANK * 8)) \
distributed_training.distributed_init_method="tcp://${MASTER_IP}:${MASTER_PORT}" \
hydra.sweep.dir=/path/to/outputs

For personal communication related to this package, please contact Linyuan Gong ([email protected]), Chenyan Xiong ([email protected]) and Xiaodong Liu ([email protected]).

FairSeq: https://github.com/pytorch/fairseq

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.

When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.