huggingface/transformers

Pull Requests Commits

Add custom_generate integration analysis to investigation notes

dacorvo committed 27 days ago

9d4bb261

Update investigation notes with full neuron_sample comparison and benchmarks

dacorvo committed 27 days ago

7ded19ec

Pre-build static 4D causal mask for decode loop in _static_sample

dacorvo committed 27 days ago

e3106956

Move output_ids and EOS masking to CPU in _static_sample

dacorvo committed 27 days ago

8ea9438f

Move EOS tracking to CPU in _static_sample to avoid blocking device sync

dacorvo committed 27 days ago

20cb3c8e

Fix left-padding position_ids in _static_sample for batched generation

dacorvo committed 27 days ago

55fc5dc9

Add investigation notes for remaining neuron_sample optimizations

dacorvo committed 27 days ago

68c6465a

Auto-dispatch to _static_sample when StaticCache is detected

dacorvo committed 27 days ago

5ea795ba

Replace scatter_ with direct indexing for output buffer in _static_sample

dacorvo committed 27 days ago

e7ea0ee9

Replace .item() device-host sync with loop index in _static_sample

dacorvo committed 27 days ago

ee8603b6

Remove redundant past_key_values reassignment in _static_sample

dacorvo committed 27 days ago

2b2bcb01

Bypass prepare_inputs_for_generation in _static_sample decode loop

dacorvo committed 27 days ago

be80b096

Add _static_sample method for static-shape generation with StaticCache

dacorvo committed 27 days ago

3dc07ec0

feat(integration): Add KubeflowCallback to enable automatic progress … (#44487)

abhijeet-dhumal committed 27 days ago

Verified c87deb21

Add GGUF support for MiniMax-M2.1 model (#44526)

JoursBleu committed 27 days ago

Verified aa57e1cd

Centralize AI agent templates in `.ai` (#44489)

tarekziade committed 27 days ago

Verified 670f3c85

support xxxFast alias in v5 tokenizers (#44766)

itazap committed 27 days ago

Verified 7cd2dd86

Remove cache_position in more models (3) (#44759)

Cyrilvallez committed 27 days ago

Verified 83a6c5b5

Fix `supports_{tp/pp}_plan` (#44696)

hmellor committed 27 days ago

Verified 9f93b612

[CI] Temporarily skip Mistral4 tests as they almost all fail (#44825)

Cyrilvallez committed 27 days ago

Verified 09fea1e6

update flex attention to use `return_aux` instead of `return_lse` when torch verison >= 2.9 (#44684)

ntenenz committed 27 days ago

Verified 2bbbbee3

[Gemma] Update conversion scripts for Transformers v5 Comaptibility (#44631)

RyanMullins committed 27 days ago

Verified 6f308258

fix bug embedding_size mismatch with hidden_size in electra model test (#44657)

kaixuanliu committed 27 days ago

Verified 3bf92002

Fix pegasus conversion (#44571)

ArthurZucker committed 27 days ago

Verified 6b01cc4b

Fix repo-check bot (#44812)

ydshieh committed 27 days ago

Verified 869dea75

Sdpa for owlvit (#42136)

Aravind-11 committed 28 days ago

Verified f1f34de0

[docs] is_causal feature (#44777)

stevhliu committed 28 days ago

Verified 03205a9d

docs(tasks): remove references to removed question-answering pipeline (#44787)

BillionClaw committed 28 days ago

Verified acc89e74

Fix configs with `@strict` (#44770)

zucchini-nlp committed 28 days ago

Verified 2f896be7

[AMD CI] Fix test failures across important models (#44632)

Abdennacer-Badaoui committed 28 days ago

Verified af93d384

Older