python run_clm_mp.py \
    	--model_name_or_path "." \
  	--tokenizer_name gpt2 \
   	--dataset_name wikitext --dataset_config_name wikitext-2-raw-v1 \
     	--do_train \
       	--do_eval  \
     	--block_size 1024 \
   	--num_train_epochs 5 \
   	--learning_rate 4e-6 \
	--per_device_train_batch_size 2 \
       	--per_device_eval_batch_size 2 \
    	--overwrite_output_dir \
	--output_dir /mnt/disks/flaxdisk/output/ \
	--cache_dir /mnt/disks/flaxdisk/cache/ \
       	--dtype bfloat16 \
   	--logging_steps 97 \
       	--eval_steps 96