# precompilation command
!MALLOC_ARENA_MAX=64 neuron_parallel_compile torchrun --nproc_per_node=32 scripts/run_clm.py \
 --model_id {model_id} \
 --dataset_path {dataset_path} \
 --bf16 True \
 --learning_rate 5e-5 \
 --output_dir dolly_llama \
 --overwrite_output_dir True \
 --per_device_train_batch_size 1 \
 --gradient_checkpointing True \
 --tensor_parallel_size 8 \
 --max_steps 10 \
 --logging_steps 10 \
 --gradient_accumulation_steps 16