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

Fix more inefficient PT operations (#37060)

Browse Source 
* Fix inefficient operations

* Remove cpu() call

* Reorder detach()

* Reorder detach()

* tolist without detach

* item without detach

* Update src/transformers/models/rag/modeling_rag.py

Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>

* Update tests/models/encodec/test_modeling_encodec.py

Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>

* Use detach().cpu().numpy

* Revert some numpy operations

* More fixes

---------

Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>
This commit is contained in:
cyyever
2025-03-31 23:31:24 +08:00
committed by GitHub
parent a1e389e637
commit 786d9c5ed9
54 changed files with 106 additions and 104 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
tests/models/mimi/test_modeling_mimi.py
View File

@@ -540,7 +540,7 @@ class MimiIntegrationTest(unittest.TestCase):
# use max bandwidth for best possible reconstruction
encoder_outputs = model.encode(inputs["input_values"], num_quantizers=int(num_codebooks))
audio_code_sums = encoder_outputs[0].sum().cpu().item()
audio_code_sums = encoder_outputs[0].sum().item()
# make sure audio encoded codes are correct
# assert relative difference less than a threshold, because `audio_code_sums` varies a bit