Cross-Thought for Sentence Encoder Pre-training (https://arxiv.org/abs/2010.03652)
git clone --depth 1 --branch v0.10.0 https://github.com/pytorch/fairseq.git
git clone https://github.com/ngoyal2707/Megatron-LM.git fairseq/fairseq/model_parallel/megatron
cp src/transformer_sentence_encoder.py fairseq/fairseq/modules/transformer_sentence_encoder.py
pip install --editable ./fairseq
Overall, we only change one file 'transformer_sentence_encoder.py' from RoBERTa for Cross-Thought. All the data processing follows RoBERTa setting.
Please follow RoBERTa finetune (https://github.com/pytorch/fairseq/blob/master/examples/roberta/README.glue.md) for data process
Cross-Thought checkpoint (https://drive.google.com/file/d/11lnZijWuRcPT07xiEO5NMhkXMfKopIb9/view?usp=sharing)
export CTPRETRAIN=False
TOTAL_NUM_UPDATES=113272 # 10 epochs through RTE for bsz 16
WARMUP_UPDATES=28318 # 6 percent of the number of updates
LR=1e-05 # Peak LR for polynomial LR scheduler.
NUM_CLASSES=2
MAX_SENTENCES=32 # Batch size.
ROBERTA_PATH=/path/to/cross-thought-checkpoint.pt
CUDA_VISIBLE_DEVICES=0 fairseq-train data/QQP-bin/ \
--restore-file $ROBERTA_PATH \
--max-positions 512 \
--batch-size $MAX_SENTENCES \
--max-tokens 4400 \
--task sentence_prediction \
--reset-optimizer --reset-dataloader --reset-meters \
--required-batch-size-multiple 1 \
--init-token 0 --separator-token 0 \
--arch roberta_base \
--criterion sentence_prediction \
--num-classes $NUM_CLASSES \
--dropout 0 --attention-dropout 0 \
--weight-decay 0.1 --optimizer adam --adam-betas "(0.9, 0.98)" --adam-eps 1e-06 \
--clip-norm 0.0 \
--lr-scheduler polynomial_decay --lr $LR --total-num-update $TOTAL_NUM_UPDATES --warmup-updates $WARMUP_UPDATES \
--fp16 --fp16-init-scale 4 --threshold-loss-scale 1 --fp16-scale-window 128 \
--max-epoch 10 \
--find-unused-parameters \
--best-checkpoint-metric accuracy --maximize-best-checkpoint-metric;
Please follow RoBERTa pre-train (https://github.com/pytorch/fairseq/blob/master/examples/roberta/README.pretraining.md) for data process
export CTPRETRAIN=True
fairseq-train --fp16 $DATA_DIR --task masked_lm --criterion masked_lm \
--arch roberta_base --sample-break-mode none --tokens-per-sample 32000 \
--optimizer adam --adam-betas '(0.9,0.98)' --adam-eps 1e-6 --clip-norm 0.0 \
--lr-scheduler polynomial_decay --lr 0.0005 --warmup-updates 10000 \
--total-num-update 125000 --dropout 0.1 --attention-dropout 0.1 --weight-decay 0.01 \
--batch-size 1 --update-freq 16 --max-update 500000 --log-format simple --log-interval 1 \
--save-dir outputs/crossthought --save-interval-updates 5000
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.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., label, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repositories 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 opencode@microsoft.com with any additional questions or comments.
MIT