File size: 884 Bytes
b221124 e0ba7eb b221124 e0ba7eb b221124 6d9b0b2 b221124 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 |
# set train hyperparams
unset LD_PRELOAD
export HF_DATASETS_CACHE="/researchdisk/datasets_cache"
export USE_TORCH=0
python3 run_clm_flax.py \
--output_dir="./" \
--model_type="gpt2" \
--config_name="./" \
--tokenizer_name="./" \
--dataset_filepath="/researchdisk/training_dataset_full_deduplicated" \
--do_train --do_eval \
--block_size="512" \
--per_device_train_batch_size="8" \
--per_device_eval_batch_size="8" \
--preprocessing_num_workers="96" \
--adam_beta1="0.9" \
--adam_beta2="0.98" \
--learning_rate="4e-5" \
--weight_decay="0.01" \
--warmup_steps="4000" \
--cosine_decay \
--overwrite_output_dir \
--logging_steps="500" \
--eval_steps="10000" \
--save_steps="10000" \
--num_train_epochs="5" \
--dtype="bfloat16" \
--push_to_hub \
--hub_model_id="Finnish-NLP/gpt2-large-finnish" |