vLLM/vllm/model_executor/models at 583a00257d4c5d1a54063d956057df1df6822b06 - vLLM

transformers

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

__init__.py

[Bugfix][Model Runner V2] Restore multimodal draft capability detection (#50417 )

2026-08-03 14:18:14 -07:00

adapters.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[aA][bB]" (#48977 )

2026-08-06 19:30:15 +00:00

afmoe.py

[Refactor] Delete dead code in models (#51838 )

2026-08-11 15:44:26 -04:00

aimv2.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[aA][bB]" (#48977 )

2026-08-06 19:30:15 +00:00

apertus.py

[Perf] RMSNorm uncontiguous support, 1.2~3.1x kernel performance improvement (#49750 )

2026-07-29 13:45:38 -07:00

arcee.py

…

arctic.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

aria.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[aA][bB]" (#48977 )

2026-08-06 19:30:15 +00:00

audioflamingo3.py

[Core] Check for GPU<->CPU syncs during CI (#43107 )

2026-08-14 17:42:05 -07:00

AXK1.py

[MOE] Standardize and abstract fused shared expert optimization selection (#51695 )

2026-08-18 12:56:56 -04:00

bagel.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[aA][bB]" (#48977 )

2026-08-06 19:30:15 +00:00

bailing_moe_linear.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[aA][bB]" (#48977 )

2026-08-06 19:30:15 +00:00

bailing_moe_mtp.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[aA][bB]" (#48977 )

2026-08-06 19:30:15 +00:00

bailing_moe_v3_mtp.py

[Model] [Quantization] Add Ling hybrid MXFP4 routed experts support (#52114 )

2026-08-13 21:46:53 +08:00

bailing_moe_v3.py

[Model] [Quantization] Add Ling hybrid MXFP4 routed experts support (#52114 )

2026-08-13 21:46:53 +08:00

bailing_moe.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

bee.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[aA][bB]" (#48977 )

2026-08-06 19:30:15 +00:00

bert_with_rope.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[aA][bB]" (#48977 )

2026-08-06 19:30:15 +00:00

bert.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[aA][bB]" (#48977 )

2026-08-06 19:30:15 +00:00

blip2.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[aA][bB]" (#48977 )

2026-08-06 19:30:15 +00:00

blip.py

…

bloom.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

chameleon.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

chatglm.py

…

cheers.py

…

clip.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[cC][dD]" (#52003 )

2026-08-13 11:33:55 +08:00

cohere2_moe.py

[Cohere] Misc changes to cohere model definitions (#50156 )

2026-08-18 04:59:57 +00:00

cohere2_vision.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

cohere_asr.py

[Core] Check for GPU<->CPU syncs during CI (#43107 )

2026-08-14 17:42:05 -07:00

cohere_eagle.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[cC][dD]" (#52003 )

2026-08-13 11:33:55 +08:00

colbert.py

…

colmodernvbert.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[cC][dD]" (#52003 )

2026-08-13 11:33:55 +08:00

colpali.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[cC][dD]" (#52003 )

2026-08-13 11:33:55 +08:00

colqwen3_5.py

[EPD] Allow KV consumers to omit MM embeddings (#52697 )

2026-08-19 02:16:08 +00:00

colqwen3.py

[EPD] Allow KV consumers to omit MM embeddings (#52697 )

2026-08-19 02:16:08 +00:00

commandr.py

[Cohere] Misc changes to cohere model definitions (#50156 )

2026-08-18 04:59:57 +00:00

config.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[cC][dD]" (#52003 )

2026-08-13 11:33:55 +08:00

conformer_encoder.py

…

cosmos3_edge.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[cC][dD]" (#52003 )

2026-08-13 11:33:55 +08:00

cosmos3.py

Cosmos3 FP8 ModelOpt/Diffusers remapping (#48952 )

2026-07-20 11:31:48 -07:00

dbrx.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[cC][dD]" (#52003 )

2026-08-13 11:33:55 +08:00

deepencoder2.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[cC][dD]" (#52003 )

2026-08-13 11:33:55 +08:00

deepencoder.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[cC][dD]" (#52003 )

2026-08-13 11:33:55 +08:00

deepseek_eagle3.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[cC][dD]" (#52003 )

2026-08-13 11:33:55 +08:00

deepseek_eagle.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[cC][dD]" (#52003 )

2026-08-13 11:33:55 +08:00

deepseek_mtp.py

[MOE] Standardize and abstract fused shared expert optimization selection (#51695 )

2026-08-18 12:56:56 -04:00

deepseek_ocr2.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

deepseek_ocr.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

deepseek_v2.py

[MOE] Standardize and abstract fused shared expert optimization selection (#51695 )

2026-08-18 12:56:56 -04:00

deepseek_vl2.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

diffusion_gemma.py

[Platform] Fill in the missing backend parameter for torch.compile (#51781 )

2026-08-19 10:31:19 +00:00

dots_ocr.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[cC][dD]" (#52003 )

2026-08-13 11:33:55 +08:00

eagle2_5_vl.py

…

ernie45_moe.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

ernie45_vl_moe.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

ernie45_vl.py

[BugFix] Revert incorrect MM keep_on_cpu=True changes (#52881 )

2026-08-19 07:07:40 +00:00

ernie45.py

…

ernie_mtp.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

exaone4_5_mtp.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

exaone4_5.py

…

exaone4.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

exaone_moe_mtp.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

exaone_moe.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

exaone.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

extract_hidden_states.py

…

fairseq2_llama.py

…

falcon_h1.py

…

falcon.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

fireredasr2.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

fireredlid.py

[BugFix] Revert incorrect MM keep_on_cpu=True changes (#52881 )

2026-08-19 07:07:40 +00:00

flex_olmo.py

[MoE Refactor] Rename FusedMoE to FusedMoEFactory (#44941 )

2026-07-31 11:28:18 +01:00

funasr.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

funaudiochat.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

gemma2.py

…

gemma3_mm.py

…

gemma3.py

[Bugfix] Default Gemma3 Model intermediate_tensors to None (#50777 )

2026-08-03 15:07:59 -04:00

gemma3n_audio_utils.py

…

gemma3n_mm.py

[Perf] Avoid some more unnecessary GPU<->CPU syncs (#51458 )

2026-08-08 17:13:26 -07:00

gemma3n.py

Remove even more unnecessary load_weights methods (#48496 )

2026-07-18 08:40:27 +00:00

gemma4_dspark.py

Fix Gemma 4 for upcoming Transformers version (#49797 )

2026-08-10 07:48:26 +00:00

gemma4_mm.py

[Bugfix][Multimodal] Keep Gemma 4 video frame counts on CPU (#52441 )

2026-08-16 09:17:04 +00:00

gemma4_mtp.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

gemma4_unified.py

[Bugfix][Model] Gemma3n/Gemma4: pad variable-length audio batches (#50958 )

2026-08-05 12:46:00 +08:00

gemma4.py

Fix Gemma 4 for upcoming Transformers version (#49797 )

2026-08-10 07:48:26 +00:00

gemma.py

…

glm4_1v.py

[Perf] Avoid more GPU<->CPU syncs in multimodal encoders (#52369 )

2026-08-14 10:40:14 -07:00

glm4_moe_lite_mtp.py

[MOE] Standardize and abstract fused shared expert optimization selection (#51695 )

2026-08-18 12:56:56 -04:00

glm4_moe_lite.py

[MOE] Standardize and abstract fused shared expert optimization selection (#51695 )

2026-08-18 12:56:56 -04:00

glm4_moe_mtp.py

[MOE] Standardize and abstract fused shared expert optimization selection (#51695 )

2026-08-18 12:56:56 -04:00

glm4_moe.py

[MOE] Standardize and abstract fused shared expert optimization selection (#51695 )

2026-08-18 12:56:56 -04:00

glm4.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

glm4v.py

…

glm_ocr_mtp.py

Remove even more unnecessary load_weights methods (#48496 )

2026-07-18 08:40:27 +00:00

glm_ocr.py

[Refactor] Remove multiple dead codes (#50285 )

2026-08-03 10:59:08 -04:00

glm.py

…

glmasr_utils.py

…

glmasr.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

gpt2.py

Remove even more unnecessary load_weights methods (#48496 )

2026-07-18 08:40:27 +00:00

gpt_j.py

…

gpt_neox.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

gpt_oss.py

[MoE Refactor] Rename FusedMoE to FusedMoEFactory (#44941 )

2026-07-31 11:28:18 +01:00

granite4_vision.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

granite_speech_plus.py

…

granite_speech.py

[Core] Check for GPU<->CPU syncs during CI (#43107 )

2026-08-14 17:42:05 -07:00

granite.py

[Model] Add GraniteSWA and GraniteMoeSWA via existing Granite (#52706 )

2026-08-18 20:32:46 -07:00

granitemoe.py

[Model] Add GraniteSWA and GraniteMoeSWA via existing Granite (#52706 )

2026-08-18 20:32:46 -07:00

granitemoehybrid.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

granitemoeshared.py

[Model] Add GraniteSWA and GraniteMoeSWA via existing Granite (#52706 )

2026-08-18 20:32:46 -07:00

gritlm.py

…

h2ovl.py

…

hrm_text.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

hunyuan_v1.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

hunyuan_vision.py

[EPD] Allow KV consumers to omit MM embeddings (#52697 )

2026-08-19 02:16:08 +00:00

hy_v3_mtp.py

Remove even more unnecessary load_weights methods (#48496 )

2026-07-18 08:40:27 +00:00

hy_v3.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

hyperclovax_vision_v2.py

…

hyperclovax_vision.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

hyperclovax.py

[ROCm][CI] fix test_common.py (#48676 )

2026-07-15 06:42:06 +00:00

idefics2_vision_model.py

[Core] Check for GPU<->CPU syncs during CI (#43107 )

2026-08-14 17:42:05 -07:00

idefics3.py

[Core] Check for GPU<->CPU syncs during CI (#43107 )

2026-08-14 17:42:05 -07:00

interfaces_base.py

…

interfaces.py

[K3] support recoverssm for K3 (#51855 )

2026-08-17 19:34:14 +08:00

intern_vit.py

Remove even more unnecessary load_weights methods (#48496 )

2026-07-18 08:40:27 +00:00

internlm2.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

interns1_pro.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

interns1_vit.py

Remove even more unnecessary load_weights methods (#48496 )

2026-07-18 08:40:27 +00:00

interns1.py

[Perf] Avoid more GPU<->CPU syncs in multimodal encoders (#52369 )

2026-08-14 10:40:14 -07:00

interns2_mobius.py

[Bugfix][Refactor] Keep Qwen3Next layer boundaries sequence parallel (#50685 )

2026-08-14 11:58:32 +08:00

interns2_preview.py

…

internvl.py

[Perf] Avoid more GPU<->CPU syncs on the model execution path (#51738 )

2026-08-11 19:30:34 -07:00

iquest_loopcoder.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

isaac.py

[BugFix] Revert incorrect MM keep_on_cpu=True changes (#52881 )

2026-08-19 07:07:40 +00:00

jais2.py

…

jamba.py

[MoE Refactor] Rename FusedMoE to FusedMoEFactory (#44941 )

2026-07-31 11:28:18 +01:00

jina_vl.py

…

jina.py

[Core] Check for GPU<->CPU syncs during CI (#43107 )

2026-08-14 17:42:05 -07:00

kanana_v.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

keye_vl1_5.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

keye.py

[EPD] Allow KV consumers to omit MM embeddings (#52697 )

2026-08-19 02:16:08 +00:00

kimi_audio.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

kimi_k25_vit.py

[Platform] Fill in the missing backend parameter for torch.compile (#51781 )

2026-08-19 10:31:19 +00:00

kimi_k25.py

[MM][CG] Support ViT full CUDA graph for Kimi-K2.5 (#50929 )

2026-08-04 17:15:56 +00:00

kimi_vl.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

laguna_dflash.py

…

laguna.py

[Refactor] Delete dead code in models (#51838 )

2026-08-11 15:44:26 -04:00

lfm2_moe.py

[Refactor] Delete dead code in models (#51838 )

2026-08-11 15:44:26 -04:00

lfm2_siglip2.py

…

lfm2_vl.py

…

lfm2.py

[Bugfix] Fix LFM2 ShortConv prefix breaking quant ignore list (#51495 )

2026-08-08 11:27:13 +00:00

lightonocr.py

…

llama4_eagle.py

[Bugfix][Model Runner V2] Restore multimodal draft capability detection (#50417 )

2026-08-03 14:18:14 -07:00

llama4.py

[MoE Refactor] Rename FusedMoE to FusedMoEFactory (#44941 )

2026-07-31 11:28:18 +01:00

llama_eagle3.py

[Bugfix][Spec Decode] Fix eagle3 first-layer qkv_proj prefix for quantized drafts (#48068 )

2026-07-16 11:34:41 +08:00

llama_eagle.py

…

llama.py

[Mypy Fix] Mypy fix for "vllm/model_executor/models/[aA][bB]" (#48977 )

2026-08-06 19:30:15 +00:00

llava_next_video.py

…

llava_next.py

[Model] Enable LoRA support for tower and connector in LlavaNextForConditionalGeneration (#49788 )

2026-08-18 15:53:49 +00:00

llava_onevision2.py

[BugFix] Revert incorrect MM keep_on_cpu=True changes (#52881 )

2026-08-19 07:07:40 +00:00

llava_onevision.py

…

llava.py

…

longcat_flash_mtp.py

…

longcat_flash_ngram.py

[ModelRunnerV2] Support prompt embeds (#42963 )

2026-08-17 21:58:50 +00:00

longcat_flash.py

[MoE Refactor] Rename FusedMoE to FusedMoEFactory (#44941 )

2026-07-31 11:28:18 +01:00

mamba2.py

…

mamba.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

medusa.py

…

mellum.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

midashenglm.py

…

mimo_audio.py

…

mimo_mtp.py

…

mimo_v2_mtp.py

…

mimo_v2_omni.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

mimo_v2.py

[Refactor] Delete dead code in models (#51838 )

2026-08-11 15:44:26 -04:00

mimo.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

minicpm3.py

[Perf] RMSNorm uncontiguous support, 1.2~3.1x kernel performance improvement (#49750 )

2026-07-29 13:45:38 -07:00

minicpm_eagle.py

…

minicpm.py

…

minicpmo.py

…

minicpmv4_6.py

[BugFix] Revert incorrect MM keep_on_cpu=True changes (#52881 )

2026-08-19 07:07:40 +00:00

minicpmv.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

minimax_m2.py

[MoE Refactor] Rename FusedMoE to FusedMoEFactory (#44941 )

2026-07-31 11:28:18 +01:00

mistral3.py

…

mistral_eagle.py

[Bugfix][Model Runner V2] Restore multimodal draft capability detection (#50417 )

2026-08-03 14:18:14 -07:00

mistral_large_3_eagle.py

[Bugfix][Model Runner V2] Restore multimodal draft capability detection (#50417 )

2026-08-03 14:18:14 -07:00

mistral_large_3.py

…

mistral.py

…

mixtral.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

mllama4.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

mlp_speculator.py

…

modernbert.py

…

module_mapping.py

…

molmo2.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

molmo.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

moondream3.py

…

moonvit.py

[Core] Check for GPU<->CPU syncs during CI (#43107 )

2026-08-14 17:42:05 -07:00

moss_audio.py

[Core] Check for GPU<->CPU syncs during CI (#43107 )

2026-08-14 17:42:05 -07:00

moss_transcribe_diarize.py

[BugFix] Revert incorrect MM keep_on_cpu=True changes (#52881 )

2026-08-19 07:07:40 +00:00

mpt.py

…

muse_glimmer.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

nano_nemotron_vl.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

nemotron_h_mtp.py

…

nemotron_h.py

[Refactor] Delete dead code in models (#51838 )

2026-08-11 15:44:26 -04:00

nemotron_nas.py

…

nemotron_parse.py

[Refactor] Remove multiple dead codes (#50285 )

2026-08-03 10:59:08 -04:00

nemotron_vl.py

…

nemotron.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

nvlm_d.py

…

olmo3.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

olmo_hybrid.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

olmoe.py

[MoE Refactor] Rename FusedMoE to FusedMoEFactory (#44941 )

2026-07-31 11:28:18 +01:00

openai_privacy_filter.py

…

opencua.py

[EPD] Allow KV consumers to omit MM embeddings (#52697 )

2026-08-19 02:16:08 +00:00

openpangu_mtp.py

…

openpangu_vl.py

…

openpangu.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

openvla.py

…

opt.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

orion.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

ovis2_5.py

[Bugfix] Fix Ovis2_5 special tokens for transformers v5 (#47298 )

2026-07-21 08:09:20 +00:00

ovis.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

paddleocr_vl.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

paligemma.py

[Bugfix][PaliGemma] Remove stale image embedding scaling (#52692 )

2026-08-18 16:00:55 +00:00

parakeet.py

[Platform] Fill in the missing backend parameter for torch.compile (#51781 )

2026-08-19 10:31:19 +00:00

param2moe.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

phi3.py

…

phi3v.py

[Perf] Avoid more GPU<->CPU syncs on the model execution path (#51738 )

2026-08-11 19:30:34 -07:00

phi4mm_audio.py

…

phi4mm_utils.py

[Refactor] Remove multiple dead codes (#50285 )

2026-08-03 10:59:08 -04:00

phi4mm.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

phi4siglip.py

…

phi.py

…

phimoe.py

[MoE Refactor] Rename FusedMoE to FusedMoEFactory (#44941 )

2026-07-31 11:28:18 +01:00

pixtral.py

…

plamo3.py

…

qianfan_ocr.py

…

qwen2_5_omni_thinker.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

qwen2_5_vl.py

[Core] Check for GPU<->CPU syncs during CI (#43107 )

2026-08-14 17:42:05 -07:00

qwen2_audio.py

[BugFix] Revert incorrect MM keep_on_cpu=True changes (#52881 )

2026-08-19 07:07:40 +00:00

qwen2_moe.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

qwen2_rm.py

…

qwen2_vl.py

[EPD] Allow KV consumers to omit MM embeddings (#52697 )

2026-08-19 02:16:08 +00:00

qwen2.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

qwen3_5_mtp.py

[MOE] Standardize and abstract fused shared expert optimization selection (#51695 )

2026-08-18 12:56:56 -04:00

qwen3_5.py

[Bugfix][LoRA] Add embedding_modules for Qwen3.5 CausalLM (#48850 )

2026-08-19 04:55:20 +00:00

qwen3_asr_forced_aligner.py

…

qwen3_asr_realtime.py

…

qwen3_asr.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

qwen3_dflash.py

Add Muse Glimmer model support (#51655 )

2026-08-14 12:28:56 +08:00

qwen3_dspark.py

[Spec Decode] DSpark confidence-scheduled verification (#47808 )

2026-08-12 12:27:48 -07:00

qwen3_eagle3.py

…

qwen3_moe.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

qwen3_next_mtp.py

[MOE] Standardize and abstract fused shared expert optimization selection (#51695 )

2026-08-18 12:56:56 -04:00

qwen3_next.py

[MOE] Standardize and abstract fused shared expert optimization selection (#51695 )

2026-08-18 12:56:56 -04:00

qwen3_omni_moe_thinker.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

qwen3_vl_moe.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

qwen3_vl.py

[EPD] Allow KV consumers to omit MM embeddings (#52697 )

2026-08-19 02:16:08 +00:00

qwen3.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

radio.py

[Refactor] Remove dead code in multiple files (#49745 )

2026-07-27 15:58:26 -04:00

registry.py

Add NemotronH_Omni_Reasoning_V3 as a supported Nemotron architecture (#52929 )

2026-08-19 12:34:50 +00:00

rnj1.py

…

roberta.py

[Model] Add RobertaForTokenClassification / XLMRobertaForTokenClassification (#47991 )

2026-07-15 14:30:33 +00:00

rvl.py

…

sarvam.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

seed_oss.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

siglip2navit.py

…

siglip.py

[Core] Check for GPU<->CPU syncs during CI (#43107 )

2026-08-14 17:42:05 -07:00

skyworkr1v.py

[Refactor] Remove multiple dead codes (#50285 )

2026-08-03 10:59:08 -04:00

smolvlm.py

…

solar.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

stablelm.py

Standardise weight tying on ParallelLMHead.tie_weights (#52147 )

2026-08-13 15:10:09 +00:00

step1.py

…

step3_text.py

[MoE Refactor] Rename FusedMoE to FusedMoEFactory (#44941 )

2026-07-31 11:28:18 +01:00

step3_vl.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

step3p5_mtp.py

fix(step3p5-mtp): honor exclude_modules for the MTP head via prefix (#48883 )

2026-07-29 16:58:29 +00:00

step3p5.py

[Refactor] Delete dead code in models (#51838 )

2026-08-11 15:44:26 -04:00

step3p7.py

…

step_vl.py

…

telechat2.py

Remove even more unnecessary load_weights methods (#48496 )

2026-07-18 08:40:27 +00:00

teleflm.py

…

terratorch.py

[Core] Check for GPU<->CPU syncs during CI (#43107 )

2026-08-14 17:42:05 -07:00

ultravox.py

[MM] Keep more metadata tensors on CPU (#52827 )

2026-08-18 19:56:33 -07:00

unlimited_ocr.py

…

utils.py

[Refactor] Remove multiple dead codes (#50285 )

2026-08-03 10:59:08 -04:00

vision.py

replace batch_norm to numerically identical without cudnn (#51734 )

2026-08-11 12:43:20 +08:00

voxtral_realtime.py

…

voxtral.py

[Perf] Avoid some more unnecessary GPU<->CPU syncs (#51458 )

2026-08-08 17:13:26 -07:00

voyage.py

…

whisper_causal.py

…

whisper_utils.py

…

whisper.py

[Model][LoRA] Add tower/connector LoRA support for Ultravox (#48215 )

2026-08-12 22:57:59 -07:00

zamba2.py

…