|
export MODEL_DIR=/home/Z6HJB/train-roberta-base-danish/roberta-base-danish/ |
|
source /home/Z6HJB/test/bin/activate |
|
|
|
python3 ./src/danish_run_mlm_flax_stream.py \ |
|
--model_name_or_path="${MODEL_DIR}" \ |
|
--output_dir="${MODEL_DIR}" \ |
|
--tokenizer_name="${MODEL_DIR}" \ |
|
--dataset_name="mc4" \ |
|
--dataset_config_name="unshuffled_deduplicated_en" \ |
|
--max_seq_length="128" \ |
|
--per_device_train_batch_size="128" \ |
|
--per_device_eval_batch_size="128" \ |
|
--learning_rate="1.087e-6" \ |
|
--warmup_steps="1000" \ |
|
--overwrite_output_dir \ |
|
--adam_beta1="0.9" \ |
|
--adam_beta2="0.98" \ |
|
--num_train_steps="200000" \ |
|
--num_eval_samples="5000" \ |
|
--save_steps="1000" \ |
|
--logging_steps="250" \ |
|
--eval_steps="1000" \ |
|
|
|
|
|
|
|
|
