diff --git a/train_mm_zero3_lora.sh b/train_mm_zero3_lora.sh index 6a14c3a..5a6d868 100755 --- a/train_mm_zero3_lora.sh +++ b/train_mm_zero3_lora.sh @@ -12,9 +12,9 @@ FORCE_COLOR=1 deepspeed --hostfile hostfile \ --warmup_ratio 0.03 \ --lora_r 16 --lora_alpha 64 --lora_dropout 0.05 \ --lora_exclude lm_head \ - --max_steps 3000 \ - --log_interval 10 \ - --eval_steps 50 \ + --max_steps 30 \ + --log_interval 1 \ + --eval_steps 5 \ --gradient_checkpointing \ --deepspeed /home/test/jd_train/ds_config_zero3_lora.json \ --report_to wandb --wandb_project ds-qwen3-lora