huggingface/transformers

Pull Requests Commits

nemotron_h_sparse: support optional latent MoE projection

ArthurZucker committed 67 days ago

84a2be05

ci: fix code-quality, repo-consistency, and doc-build for nemotron_h split

ArthurZucker committed 110 days ago

e57b10e7

test: migrate nemotron_h_{dense,sparse} tests to CausalLMModelTester

ArthurZucker committed 110 days ago

3e71a29d

Merge remote-tracking branch 'origin/main' into nemotron-h-split-dense-sparse

ArthurZucker committed 110 days ago

af3d5b0b

refactor: Jamba-style decoder layers, inherit NemotronMLP/JambaModel

ArthurZucker committed 110 days ago

da5a518f

[nemotron_h] Add support for MLP mixers (#44763)

xenova committed 110 days ago

Verified 77c0e6e7

refactor: use DecoderLayer pattern; drop moe_latent_size; inherit MixtralExperts

ArthurZucker committed 110 days ago

9907265b

ArthurZucker committed 110 days ago

8390498a

add expert parallelism for gemma-4-26B-A4B-it (#45279)

sywangyi committed 110 days ago

Verified e737fb86

Add full GGUF loading support for GPT‑OSS (fixes #43366, supersedes #43757) latest (#45506)

sirzechs66 committed 110 days ago

Verified 5e238684

Update Gemma4 weight conversion script (#45328)

RyanMullins committed 110 days ago

Verified 2da596b0

Move some conversion mappings to PrefixChange (#45567)

Cyrilvallez committed 110 days ago

Verified 71d1a7b7

fix table update versions (#45544)

tarekziade committed 110 days ago

Verified f51cf506

Add disable_mmap kwarg to from_pretrained with hf-mount auto-detection (#45547)

rtrompier committed 110 days ago

Verified 8fb7c7e5

ArthurZucker committed 110 days ago

85a0cc75

fix(DSV3): parity between native `DeepseekV3MoE` and remote official implementation (#45441)

casinca committed 110 days ago

Verified cb0adddc

Align gemma3n cache sharing to gemma4 (#45489)

Cyrilvallez committed 110 days ago

Verified 08244b96

[modular] Fix modular logic broken in #45045 (#45539)

Cyrilvallez committed 110 days ago

Verified f048e845

Fix: propagate quantization_config to text sub-config for composite models in AutoModelForCausalLM (#45494)

lvliang-intel committed 111 days ago

Verified 85099df9

T5Gemma2: fix `prepare_decoder_input_ids_from_labels` (#45516)

Tokarak committed 111 days ago

Verified fcc22811

[Trainer] Add ddp_static_graph option (#45519)

KeitaW committed 111 days ago

Verified 26bb358e

Add dtype config options for Four Over Six (#45367)

jackcook committed 111 days ago

Verified 1a4bc140

[Sam3LiteText] Remove unnecessary modules/configs (#45535)

yonigozlan committed 111 days ago

Verified 7439ac04

Fix conditional check for float formatting (#44425)

qgallouedec committed 111 days ago

Verified dadd30b8

Fix AMD CI: rebuild torchvision with libjpeg + refresh expectations (#45533)

Abdennacer-Badaoui committed 111 days ago

Verified 67fb8bb3

ArthurZucker committed 111 days ago

f4eb5e76

ArthurZucker committed 111 days ago

30bffaef

ArthurZucker committed 111 days ago

55aa51ff

Merge branch 'main' of github.com:huggingface/transformers into nemotron-h-fixes

ArthurZucker committed 111 days ago

c7acbb11

Fix EP: RouterParallel shape, tp_plan property, grouped_mm sentinels (#45473)

AmineDiro committed 111 days ago

Verified 9dff7ca5

Older