# 8 * 65GiB. 80s/it
PYTORCH_CUDA_ALLOC_CONF='expandable_segments:True' \
NPROC_PER_NODE=8 \
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
megatron sft \
    --model Qwen/Qwen2.5-72B-Instruct \
    --save_safetensors true \
    --dataset 'liucong/Chinese-DeepSeek-R1-Distill-data-110k-SFT' \
    --load_from_cache_file true \
    --split_dataset_ratio 0.01 \
    --tensor_model_parallel_size 8 \
    --micro_batch_size 1 \
    --global_batch_size 16 \
    --packing true \
    --recompute_granularity full \
    --recompute_method uniform \
    --recompute_num_layers 1 \
    --num_train_epochs 5 \
    --finetune true \
    --cross_entropy_loss_fusion true \
    --lr 1e-5 \
    --lr_warmup_fraction 0.05 \
    --min_lr 1e-6 \
    --output_dir megatron_output/Qwen2.5-72B-Instruct \
    --eval_steps 500 \
    --save_steps 500 \
    --max_length 8192 \
    --dataloader_num_workers 8 \
    --dataset_num_proc 8 \
    --no_save_optim true \
    --no_save_rng true \
    --sequence_parallel true \
    --optimizer_cpu_offload true \
    --use_precision_aware_optimizer true \
    --optimizer_offload_fraction 1 \
    --attention_backend flash