yuantao 315e3f654a Refactor code, make attn backend focus on diffkv and move sink logic to GPUModelRunner ...

Signed-off-by: yuantao <2422264527@qq.com>

2025-11-26 11:34:23 +08:00

..

transformers

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

__init__.py

…

adapters.py

[Docs] Update the name of Transformers backend -> Transformers modeling backend (#28725)

2025-11-14 16:34:14 +00:00

afmoe.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

aimv2.py

[Models] Replace all nn.Conv2d with vLLM's Conv2dLayer (#28842)

2025-11-18 18:56:04 +00:00

apertus.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

arcee.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

arctic.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

aria.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

aya_vision.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

baichuan.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

bailing_moe.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

bamba.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

bee.py

[Model][VLM] Support Bee-8B Model (#27012)

2025-10-20 02:31:26 +00:00

bert_with_rope.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

bert.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

blip2.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

blip.py

[Models] Replace all nn.Conv2d with vLLM's Conv2dLayer (#28842)

2025-11-18 18:56:04 +00:00

bloom.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

chameleon.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

chatglm.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

clip.py

[Model][MM] Extract conv layer as CustomOp (#28455)

2025-11-14 19:16:13 +08:00

cohere2_vision.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

commandr.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

config.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

dbrx.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

deepencoder.py

[Models] Replace all nn.Conv2d with vLLM's Conv2dLayer (#28842)

2025-11-18 18:56:04 +00:00

deepseek_eagle.py

[Bugfix] Fix spec decode memory regression after #28549 (#28819)

2025-11-20 19:05:50 +08:00

deepseek_mtp.py

[Bugfix][Rocm] Fix shared expert weight loading failure in DeepSeek-MTP (#27563)

2025-11-24 10:16:52 +00:00

deepseek_ocr.py

[BugFix] DeepSeek-OCR: apply NoRepeatNGramLogitsProcessor to greedy path (#28617)

2025-11-13 15:24:35 +00:00

deepseek_v2.py

[Bugfix][Rocm] Fix shared expert weight loading failure in DeepSeek-MTP (#27563)

2025-11-24 10:16:52 +00:00

deepseek_vl2.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

dots1.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

dots_ocr.py

[Core] Deprecate xformers (#29262)

2025-11-24 04:18:55 +00:00

ernie45_moe.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

ernie45_vl_moe.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

ernie45_vl.py

[Core] Deprecate xformers (#29262)

2025-11-24 04:18:55 +00:00

ernie45.py

[Model] Add PaddleOCR-VL Model Support (#27758)

2025-11-03 19:04:22 +08:00

ernie_mtp.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

exaone4.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

exaone.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

fairseq2_llama.py

…

falcon_h1.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

falcon.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

flex_olmo.py

…

fuyu.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

gemma2.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

gemma3_mm.py

[Bugfix] Revert custom attention mask for gemma3-mm (#28995)

2025-11-20 13:23:22 +08:00

gemma3.py

[CI Failure] Fix Gemma3 RoPE configuration for sliding attention layers (#29111)

2025-11-20 23:53:30 -08:00

gemma3n_mm.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

gemma3n.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

gemma.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

glm4_1v.py

[Core] Deprecate xformers (#29262)

2025-11-24 04:18:55 +00:00

glm4_moe_mtp.py

[BugFix] Fix glm4_moe_mtp load weights bug (#28805)

2025-11-17 17:13:11 +08:00

glm4_moe.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

glm4.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

glm4v.py

[Models] Replace all nn.Conv2d with vLLM's Conv2dLayer (#28842)

2025-11-18 18:56:04 +00:00

glm.py

…

gpt2.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

gpt_bigcode.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

gpt_j.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

gpt_neox.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

gpt_oss.py

[LoRA] Optimize 3D MoE logic (#29222)

2025-11-24 10:27:23 +08:00

granite_speech.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

granite.py

[LoRA][2/2]Remove LoRA extra vocab (#28545)

2025-11-21 09:46:43 +08:00

granitemoe.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

granitemoehybrid.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

granitemoeshared.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

gritlm.py

…

grok1.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

h2ovl.py

…

hunyuan_v1.py

[Model] Add HunyuanOCR support (#29327)

2025-11-25 03:28:51 +00:00

hunyuan_vision.py

[Model] Add HunyuanOCR support (#29327)

2025-11-25 03:28:51 +00:00

hyperclovax_vision.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

idefics2_vision_model.py

[Models] Replace all nn.Conv2d with vLLM's Conv2dLayer (#28842)

2025-11-18 18:56:04 +00:00

idefics3.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

interfaces_base.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

interfaces.py

[Model] Add HunyuanOCR support (#29327)

2025-11-25 03:28:51 +00:00

intern_vit.py

[Models] Replace all nn.Conv2d with vLLM's Conv2dLayer (#28842)

2025-11-18 18:56:04 +00:00

internlm2_ve.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

internlm2.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

interns1_vit.py

[Models] Replace all nn.Conv2d with vLLM's Conv2dLayer (#28842)

2025-11-18 18:56:04 +00:00

interns1.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

internvl.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

jais.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

jamba.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

jina_vl.py

…

keye_vl1_5.py

[Model] Pass mm_features directly into get_mrope_input_positions (#28399)

2025-11-11 21:14:48 +08:00

keye.py

[Core] Deprecate xformers (#29262)

2025-11-24 04:18:55 +00:00

kimi_linear.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

kimi_vl.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

lfm2_moe.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

lfm2.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

lightonocr.py

[Model] Add support for LightOnOCR (#26916)

2025-10-17 05:05:24 +00:00

llama4_eagle.py

[Bugfix] Fix spec decode memory regression after #28549 (#28819)

2025-11-20 19:05:50 +08:00

llama4.py

[BugFix] Fix Llama4 Pipeline Parallelism Assert Error (#28577)

2025-11-20 02:52:36 -08:00

llama_eagle3.py

[Spec Decode] Add support for EAGLE3 heads that do not use_aux_hidden_states (#27688)

2025-11-24 20:13:12 -05:00

llama_eagle.py

[Bugfix] Fix spec decode memory regression after #28549 (#28819)

2025-11-20 19:05:50 +08:00

llama.py

Add option to use unbacked, and backed size obl dynamic shapes for more sounds compilation. (#26199)

2025-11-24 10:12:41 -05:00

llava_next_video.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

llava_next.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

llava_onevision.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

llava.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

longcat_flash_mtp.py

…

longcat_flash.py

[Refactor] Optimize select_experts (#28069)

2025-11-19 18:53:15 -05:00

mamba2.py

[Hybrid][torch.compile] Refactor mamba2 forward to avoid obscuring linear projections under custom op (#28587)

2025-11-18 16:49:36 -08:00

mamba.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

medusa.py

[LoRA][1/N]Remove LoRA extra vocab (#28382)

2025-11-11 11:06:21 -08:00

midashenglm.py

[Models] Replace all nn.Conv2d with vLLM's Conv2dLayer (#28842)

2025-11-18 18:56:04 +00:00

mimo_mtp.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

mimo.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

minicpm3.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

minicpm_eagle.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

minicpm.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

minicpmo.py

…

minicpmv.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

minimax_m2.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

minimax_text_01.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

minimax_vl_01.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

mistral3.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

mixtral.py

[LoRA][2/2]Remove LoRA extra vocab (#28545)

2025-11-21 09:46:43 +08:00

mllama4.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

mlp_speculator.py

[LoRA][1/N]Remove LoRA extra vocab (#28382)

2025-11-11 11:06:21 -08:00

modernbert.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

module_mapping.py

…

molmo.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

moonvit.py

[XPU]fix Kimi-VL-A3B-thinking on xpu (#29309)

2025-11-24 21:02:21 +00:00

mpt.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

nano_nemotron_vl.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

nemotron_h.py

[Hybrid][torch.compile] Refactor mamba2 forward to avoid obscuring linear projections under custom op (#28587)

2025-11-18 16:49:36 -08:00

nemotron_nas.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

nemotron_vl.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

nemotron.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

nvlm_d.py

…

olmo2.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

olmo.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

olmoe.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

opencua.py

[Model] Add OpenCUA-7B support (#29068)

2025-11-24 10:27:55 +08:00

openpangu_mtp.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

openpangu.py

Refactor code, make attn backend focus on diffkv and move sink logic to GPUModelRunner

2025-11-26 11:34:23 +08:00

opt.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

orion.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

ouro.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

ovis2_5.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

ovis.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

paddleocr_vl.py

[Core] Deprecate xformers (#29262)

2025-11-24 04:18:55 +00:00

paligemma.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

persimmon.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

phi3.py

…

phi3v.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

phi4_multimodal.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

phi4mm_audio.py

…

phi4mm_utils.py

…

phi4mm.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

phi.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

phimoe.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

pixtral.py

[Core] Deprecate xformers (#29262)

2025-11-24 04:18:55 +00:00

plamo2.py

[V0 deprecation] Remove more V0 references (#29088)

2025-11-21 11:56:59 +00:00

plamo3.py

[Bugfix] Fix Plamo3 rope handling (#29092)

2025-11-21 11:38:35 +08:00

qwen2_5_omni_thinker.py

[Multimodal][Qwen3 Omni] Make Qwen3 Omni work with audio-in-video inputs in V1 engine. (#27721)

2025-11-24 19:24:37 +00:00

qwen2_5_vl.py

[Core] Deprecate xformers (#29262)

2025-11-24 04:18:55 +00:00

qwen2_audio.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

qwen2_moe.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

qwen2_rm.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

qwen2_vl.py

[Core] Deprecate xformers (#29262)

2025-11-24 04:18:55 +00:00

qwen2.py

Add option to use unbacked, and backed size obl dynamic shapes for more sounds compilation. (#26199)

2025-11-24 10:12:41 -05:00

qwen3_moe.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

qwen3_next_mtp.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

qwen3_next.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

qwen3_omni_moe_thinker.py

[Multimodal][Qwen3 Omni] Make Qwen3 Omni work with audio-in-video inputs in V1 engine. (#27721)

2025-11-24 19:24:37 +00:00

qwen3_vl_moe.py

[Feature] EPLB on Qwen3VLMoe and CompressedTensorsWNA16MoEMethod (#28849)

2025-11-19 18:30:08 -05:00

qwen3_vl.py

[Core] Deprecate xformers (#29262)

2025-11-24 04:18:55 +00:00

qwen3.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

qwen_vl.py

[Models] Replace all nn.Conv2d with vLLM's Conv2dLayer (#28842)

2025-11-18 18:56:04 +00:00

qwen.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

radio.py

Nemotron Nano V2 VL + EVS Video Support (#27107)

2025-10-20 22:19:11 +08:00

registry.py

Merge branch 'main' into Add_support_for_openpangu_promoe_v2

2025-11-25 19:33:50 +08:00

roberta.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

rvl.py

…

seed_oss.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

siglip2navit.py

[BugFix] Fix flash_attn import in siglip2navit.py (#29082)

2025-11-20 12:14:29 +00:00

siglip.py

[Models] Replace all nn.Conv2d with vLLM's Conv2dLayer (#28842)

2025-11-18 18:56:04 +00:00

skyworkr1v.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

smolvlm.py

…

solar.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

stablelm.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

starcoder2.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

step3_text.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00

step3_vl.py

[Models] Replace all nn.Conv2d with vLLM's Conv2dLayer (#28842)

2025-11-18 18:56:04 +00:00

swin.py

…

tarsier.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

telechat2.py

…

teleflm.py

[LoRA][2/2]Remove LoRA extra vocab (#28545)

2025-11-21 09:46:43 +08:00

terratorch.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

ultravox.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

utils.py

Move online quantization to model.load_weights (#26327)

2025-11-18 16:52:41 -08:00

vision.py

[Model][MM] Extract conv layer as CustomOp (#28455)

2025-11-14 19:16:13 +08:00

voxtral.py

Rename clashing method names for vLLM model protocol (#27583)

2025-11-12 19:14:33 -08:00

whisper.py

[Core] Align whisper closer to other multimodal models (#27292)

2025-11-21 12:01:54 +00:00

zamba2.py

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

2025-11-19 09:06:36 -08:00