huggingface/transformers

Pull Requests Commits

replaced the imported-module analysis path in convert_modular_file() with a metadata-free mapper visit that tracks top-level context and source order directly, eliminating expensive LibCST ParentNodeProvider/PositionProvider passes while preserving byte-identical output

tarekziade committed 1 day ago

cf602885

Python-only fast path based on symtable plus a lightweight AST pass

tarekziade committed 1 day ago

07f2a0ad

second optim / cache

tarekziade committed 1 day ago

46d70f41

investigate modular conversion speedups

tarekziade committed 1 day ago

faf20982

post release tag update

ArthurZucker committed 1 day ago

1f553bdc

style was missing sorry @ydshieh :) (#45038)

ArthurZucker committed 1 day ago

Verified 435203ec

fix: protect torch imports in processing files and fix import guards

ArthurZucker committed 1 day ago

16d437e4

Fix release full (#45029)

ArthurZucker committed 1 day ago

Verified 97b7727e

Add cohere asr (#45023)

eustlb committed 1 day ago

Verified 78bdaf0b

docs: add PermuteForRope to conversion operations reverse table (#45035)

asuryateja committed 1 day ago

Verified e8951077

[CB] Persistent manager (#44435)

remi-or committed 1 day ago

Verified 67100cca

Add BC for `_further_process_kwargs` (#45033)

hmellor committed 1 day ago

Verified 69f9d552

Use multi runners to check new failing tests in a CI run (#45032)

ydshieh committed 1 day ago

Verified e3f7cc3b

[`fix`] Use the correct _tied_weights_keys for CamembertForCausalLM (#45031)

tomaarsen committed 1 day ago

Verified 882ffdbb

change dev ver. we forgot to do this when we released 5.3.0

ArthurZucker committed 2 days ago

d81ad481

Ensure final evaluation runs with step-based evaluation strategy (#44146)

khushali9 committed 2 days ago

Verified ec196140

Add `base_model_tp_plan` to `OlmoeConfig` (#44668)

dacorvo committed 2 days ago

Verified da37a4d9

ci: add anti-slop action (#44847)

tarekziade committed 2 days ago

Verified 61efa322

Add doc page for capturing outputs (#44947)

zucchini-nlp committed 2 days ago

Verified f87d392e

Dynamic weight conversion is recursive (#44300)

zucchini-nlp committed 2 days ago

Verified 09832b2a

Don't run `tests_hub` if no tests found (#45014)

ydshieh committed 2 days ago

Verified b1ba6d7d

Fix type hint for `attention_chunk_size` in `Llama4TextConfig` (#45002)

hmellor committed 2 days ago

Verified c9faacd7

Fix AutoProcessor.from_pretrained silently dropping hub kwargs (#44710)

he-yufeng committed 2 days ago

Verified 35b005bb

Fix `maybe_autocast` crashing on meta device tensors (#44984)

Butanium committed 3 days ago

Verified c17877c2

Add VidEoMT (#44285)

NielsRogge committed 3 days ago

Verified 2a542366

fix: remove Copied from comments between @torch.jit.script and def for Python 3.13 compat (#44986)

Krishnachaitanyakc committed 3 days ago

Verified 2f624917

More small vllm fixes (#44990)

ArthurZucker committed 3 days ago

Verified 0e1978c9

fix(models): Fix Perceiver interpolate_pos_encoding interpolating to the source size (#44899)

harshaljanjani committed 3 days ago

Verified 2620c4dd

Allow `mm_token_type` be non-padded lists (#44563)

zucchini-nlp committed 3 days ago

Verified c532659b

Fix CPU 16 bytes alignment issue using equivalent fallback (#44970)

IlyasMoutawwakil committed 3 days ago

Verified c3b41c6c

Older