fix #424

Former-commit-id: daaf89f1126112a73b9f115b0f5617a8cd974a3e
2025-10-15 08:08:09 +08:00 · 2023-09-27 22:49:43 +08:00 · 2023-09-27 22:49:43 +08:00 · dd623325e8
commit dd623325e8
parent e8a375c8f2
1 changed files with 2 additions and 1 deletions
--- a/src/llmtuner/tuner/ppo/workflow.py
+++ b/src/llmtuner/tuner/ppo/workflow.py
@ -42,7 +42,8 @@ def run_ppo(
        ppo_epochs=1,
        max_grad_norm=training_args.max_grad_norm,
        seed=training_args.seed,
-        optimize_cuda_cache=True
+        optimize_cuda_cache=True,
+        accelerator_kwargs={"step_scheduler_with_optimizer": False}
    )

    if finetuning_args.ppo_score_norm: