SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Never fallback to eager implicitly (#38327)

Browse Source 
* remove arg everywhere

* Update warnings

* add more models

* Update sdpa_attention.py

* fix style

* fix

* readd warnings but not for flex

* Update test_modeling_common.py

* skip

* fix

---------

Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>
This commit is contained in:
Cyril Vallez
2025-05-23 19:48:01 +02:00
committed by GitHub
parent e64ed0304c
commit e0aad278fe
73 changed files with 66 additions and 544 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
docs/source/en/modular_transformers.md
View File

@@ -243,13 +243,7 @@ class Olmo2Attention(OlmoAttention):
attention_interface: Callable = eager_attention_forward
if self.config._attn_implementation != "eager":
if self.config._attn_implementation == "sdpa" and kwargs.get("output_attentions", False):
logger.warning_once(
"`torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to "
'eager attention. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.'
)
else:
attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
attn_output, attn_weights = attention_interface(
self,