./run_clm_flax.py \
    --output_dir="." \
    --model_type="gpt2" \
    --config_name="." \
    --tokenizer_name="." \
    --dataset_name="oscar" \
    --dataset_config_name="unshuffled_deduplicated_pl" \
    --do_train --do_eval \
    --block_size="512" \
    --per_device_train_batch_size="64" \
    --per_device_eval_batch_size="64" \
    --learning_rate="2e-4" --warmup_steps="5000" \
    --adam_beta1="0.9" --adam_beta2="0.98" --weight_decay="0.01" \
    --overwrite_output_dir \
    --num_train_epochs="3" \
    --logging_steps="3500" \
    --preprocessing_num_workers="64" \
    --save_steps="7000" \
    --eval_steps="7000" \
    --model_name_or_path="." \
    --push_to_hub \