huggingface/transformers

Pull Requests Commits

[debug] Ignore fp noise: only suppress mismatch when one side is 0.0 and other <= 1e-6

ydshieh committed 47 days ago

1e8f22f0

[debug] Lower epsilon to 1e-7 to avoid false mismatches at ~1e-6 gradient values

ydshieh committed 47 days ago

2e710a7c

[debug] Use 1e-6 epsilon for non-zero grad check to ignore fp noise from GC recomputation

ydshieh committed 47 days ago

60d4994e

Revert "[debug] Use 1e-6 epsilon for non-zero grad check to ignore fp noise from GC recomputation"

ydshieh committed 47 days ago

6301d819

[debug] Use 1e-6 epsilon for non-zero grad check to ignore fp noise from GC recomputation

ydshieh committed 47 days ago

cb297b6b

[debug] Log max grad_sum on MISMATCH for normal and gradcp runs

ydshieh committed 47 days ago

7902406c

Remove extra JOB_TO_TEST_FILE entries kept during cherry-pick

ydshieh committed 47 days ago

1362826d

Point PR CI caller to transformers-ci debug_gc branch

ydshieh committed 47 days ago

879e744d

Add missing import unittest.mock

ydshieh committed 47 days ago

6b2b93b6

ydshieh committed 47 days ago

e7b25959

Revert test_modeling_common.py to main (restore removed GC test methods)

ydshieh committed 47 days ago

d01ba351

[debug] Skip flaky gradient checkpointing reentrant tests in BigBirdPegasus

ydshieh committed 47 days ago

5794eccc

[debug] BigBirdPegasus gradient checkpointing flakiness investigation

ydshieh committed 47 days ago

e444c500

ydshieh committed 47 days ago

b80b4a39

[new model] step 3.7 (#46658)

itazap committed 57 days ago

Verified c49429ed

[Whisper] Fix batch decode_with_timestamps in WhisperTokenizer.decode() (#47997)

ydshieh committed 57 days ago

Verified 64b6595f

[Whisper] Fix integration test failures on A10G (dtype, stale values, API changes) (#47995)

ydshieh committed 57 days ago

Verified 16cbed7e

[DeepSeekV2] Fix integration tests OOM: use device_map=auto instead of 8-bit quantization (#47991)

ydshieh committed 58 days ago

Verified a61d5f9e

[OLMoE] Update expected logits for A10G and add torch.no_grad() (#47989)

ydshieh committed 58 days ago

Verified bd5df99a

[OLMo] Fix OOM in logits tests by adding torch.no_grad() (#47986)

ydshieh committed 58 days ago

Verified c7ad7287

[GPTNeoX] Fix post_processor not overridden when loading from pretrained (OLMo garbage generation) (#47988)

ydshieh committed 58 days ago

Verified ce5c8f5e

[AXK1] Fix expected logits for CUDA A10G (#47980)

ydshieh committed 59 days ago

Verified 96fe6dce

[Gemma] Update expected values for A10G (#47976)

ydshieh committed 59 days ago

Verified df0db63c

[docs] torchcodec + vision processor features (#47935)

stevhliu committed 59 days ago

Verified a597f974

Fix gemma4 video to device (#47896)

guarin committed 59 days ago

Verified c1ff1186

[emu3] 🦮 Black Labrador is back! Fix image generation broken since #37033 (#47948)

ydshieh committed 59 days ago

Verified 60eefbc8

Remove stale (None, None) fallback in qwen2_5_vl batch_different_resolutions test (#47972)

ydshieh committed 59 days ago

Verified 7acb594b

Fix CLIP _init_weights when a child module carries quantized weights (#47921)

Bluear7878 committed 59 days ago

Verified 13190137

Potential fix for code scanning alert no. 267: Artifact poisoning (#47949)

tarekziade committed 59 days ago

Verified 463f684a

Fix GatedDeltaNet A_log dtype to prevent -inf under bfloat16 init (#47944)

Nkluge-correa committed 59 days ago

Verified 95940bf8

Older