Kashif Rasul d54669a73e [Qwen] avoid creating attention masks when there is no padding (#12987) ...

* avoid creating attention masks when there is no padding

* make fix-copies

* torch compile tests

* set all ones mask to none

* fix positional encoding from becoming > 4096

* fix from review

* slice freqs_cis to match the input sequence length

* keep only attenton masking change

---------

Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>

2026-01-27 12:42:48 -10:00

..

autoencoders

Add LTX 2.0 Video Pipelines (#12915)

2026-01-07 21:24:27 -08:00

transformers

[Qwen] avoid creating attention masks when there is no padding (#12987)

2026-01-27 12:42:48 -10:00

unets

[Refactor] Move testing utils out of src (#12238)

2025-08-28 19:53:02 +05:30

__init__.py

[Tests] Improve unet / vae tests (#1018)

2022-10-28 13:43:26 +02:00

test_activations.py

move activation dispatches into helper function (#3656)

2023-06-05 12:30:48 -07:00

test_attention_processor.py

[Refactor] Move testing utils out of src (#12238)

2025-08-28 19:53:02 +05:30

test_layers_utils.py

[Refactor] Move testing utils out of src (#12238)

2025-08-28 19:53:02 +05:30

test_modeling_common.py

Fix broken group offloading with block_level for models with standalone layers (#12692)

2025-12-05 18:54:05 +05:30

test_models_auto.py

[Feature] AutoModel can load components using model_index.json (#11401)

2025-05-26 14:06:36 +05:30