From e16fcd7d947740e43fd9ee138b8761cdee2aad13 Mon Sep 17 00:00:00 2001 From: Daniel Han Date: Fri, 23 Aug 2024 23:43:57 -0700 Subject: [PATCH] Update gemma2.py --- unsloth/models/gemma2.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/unsloth/models/gemma2.py b/unsloth/models/gemma2.py index ea9f53e7db..6858f52573 100644 --- a/unsloth/models/gemma2.py +++ b/unsloth/models/gemma2.py @@ -62,7 +62,7 @@ if HAS_FLASH_ATTENTION_SOFTCAPPING: # [TODO] We must randomnly use torch.compile? # I checked the gradients and formulas and I'm sure it's correct. # I'm stumped :( -@torch.compile(fullgraph = True, dynamic = True, options = torch_compile_options) +@torch.compile(fullgraph = False, dynamic = True, options = torch_compile_options) def fast_rms_layernorm_gemma2_compiled(layernorm, X, gemma = True): old_dtype = X.dtype X = X.float()