From 9b773201f2efd0bda9b9f1cfdd23d73f12da54b7 Mon Sep 17 00:00:00 2001 From: Daniel Han Date: Fri, 14 Feb 2025 04:45:48 -0800 Subject: [PATCH] Update rl_replacements.py --- unsloth/models/rl_replacements.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/unsloth/models/rl_replacements.py b/unsloth/models/rl_replacements.py index 9b9a113f2d..36022f1e37 100644 --- a/unsloth/models/rl_replacements.py +++ b/unsloth/models/rl_replacements.py @@ -195,7 +195,7 @@ RL_FUNCTIONS["grpo_trainer"].append(grpo_trainer__get_per_token_logps) # Custom compiled GRPO loss - creates 3 Triton kernels -@torch.compile(dynamic = True, fullgraph = True, options = torch_compile_options,) +# @torch.compile(dynamic = True, fullgraph = True, options = torch_compile_options,) def _grpo_compute_loss(old_logits, new_logits, input_ids, mask, beta): old_logits = old_logits.to(torch.float32) new_logits = new_logits.to(torch.float32)