transformers/examples/research_projects/seq2seq-distillation/finetune.sh

12 lines
342 B
Bash
Executable File

# the proper usage is documented in the README, you need to specify data_dir, output_dir and model_name_or_path
# run ./finetune.sh --help to see all the possible options
python finetune.py \
--learning_rate=3e-5 \
--fp16 \
--gpus 1 \
--do_train \
--do_predict \
--n_val 1000 \
--val_check_interval 0.1 \
"$@"