diff --git a/mm-zero3.sh b/mm-zero3.sh index c8e9578..77b1858 100755 --- a/mm-zero3.sh +++ b/mm-zero3.sh @@ -1,9 +1,9 @@ deepspeed --hostfile hostfile \ --num_nodes 6 --num_gpus 4 \ /home/test/jd_train/train_sft_ds.py \ - --model_name_or_path /home/test/Qwen3-8B \ + --model_name_or_path /home/test/Qwen3-32B \ --data_glob "/home/test/datasets/my_corpus/train.jsonl" \ - --output_dir /home/test/checkpoints/q3-8b-ds4 \ + --output_dir /home/test/checkpoints/q3-32b-ds4 \ --seq_len 512 \ --per_device_train_batch_size 1 \ --gradient_accumulation_steps 1 \