ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

only warn about prepare causal attention mask when transformers<=4.52.4 (#2867)

Browse source
This commit is contained in:
DoubleMathew 2025-07-04 03:54:27 -05:00 • committed by GitHub
commit 95f4dc6137
1 changed files with 7 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

13
unsloth/models/llama.py
View file

@ -152,12 +152,13 @@ def _fast_prepare_inputs_for_generation(self, input_ids, attention_mask=None, **
)
else:
attention_mask = attention_mask[:,[-1]]
logger.warning_once(
f"{self.__class__.__name__} has no `_prepare_4d_causal_attention_mask_with_cache_position` method "
"defined in its base modeling class. Compiled forward passes will be sub-optimal. If you're "
"writing code, see Llama for an example implementation. If you're a user, please report this "
"issue on GitHub."
)
if transformers_version <= Version("4.52.4"):
logger.warning_once(
f"{self.__class__.__name__} has no `_prepare_4d_causal_attention_mask_with_cache_position` method "
"defined in its base modeling class. Compiled forward passes will be sub-optimal. If you're "
"writing code, see Llama for an example implementation. If you're a user, please report this "
"issue on GitHub."
)
if "cache_position" in kwargs:
kwargs["position_ids"] = kwargs["cache_position"]