diff --git a/examples/v1/train_full/train_full_ulysses_cp.yaml b/examples/v1/train_full/train_full_ulysses_cp.yaml index 2b7ba713a..167e3dfe6 100644 --- a/examples/v1/train_full/train_full_ulysses_cp.yaml +++ b/examples/v1/train_full/train_full_ulysses_cp.yaml @@ -3,6 +3,7 @@ trust_remote_code: true model_class: llm template: qwen3_nothink +flash_attn: flash_attention_2 # FSDP Config dist_config: