https://zenodo.org/record/5604548#.YXoG7NZBw1I
wget https://zenodo.org/record/5604548/files/train.tar.gz
wget https://zenodo.org/record/5604548/files/test.tar.gz
wget https://zenodo.org/record/5604548/files/_1_commonTags.csv
tar -xvf file.tar.gz
- Train File: ./bert/triplet/train_trinity.bash
- Test File: ./bert/triplet/test_triplet.py
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
python train_trinity.py \
--data_folder ../../data/train \
--output_dir ../../data/results \
--per_gpu_train_batch_size 2 \
--logging_steps 100 \
--gradient_accumulation_steps 4 \
--num_train_epochs 3 \
--learning_rate 1e-5 2>&1| tee train_trinity.log
distributed training
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python -m torch.distributed.launch \
--nproc_per_node=8 \
train_trinity.py \
--data_folder ../../data/train \
--output_dir ../../data/results \
--per_gpu_train_batch_size 2 \
--logging_steps 100 \
--gradient_accumulation_steps 4 \
--num_train_epochs 3 \
--learning_rate 1e-5 2>&1| tee train_trinity.log