Commits

Branch selector

 main

User selector

All users

All time

Commit History

Commits on Apr 2, 2026

• feat(token_compressor): introduce universal pruning framework and ali… (#238)

  YF-T

  authored
  0f1e8a6

  Copy full SHA for 0f1e8a6

• fix(speculative): fix qwen3/qwen2.5 (#280)

  dawnranger

  authored
  c5c8186

  Copy full SHA for c5c8186

Commits on Mar 31, 2026

• fix: AttributeError: 'Qwen3VLForConditionalGeneration' object has no attribute 'language_model' (#279)

  dawnranger

  authored
  d90d4d9

  Copy full SHA for d90d4d9

• feat(speculative): support extract hidden states with vllm backend (#273)

  dawnranger

  authored
  de4f5ec

  Copy full SHA for de4f5ec

Commits on Mar 26, 2026

• [fix]: Fix reference for weights to quant. (#276)

  Show description for b0af806
  

  C-KRSW

  and

  krizaltang

  authored
  b0af806

  Copy full SHA for b0af806

Commits on Mar 25, 2026

• [ref]: Refeactor weight-only quantization. (#275)

  Show description for 8f66b65
  

  C-KRSW

  and

  krizaltang

  authored
  8f66b65

  Copy full SHA for 8f66b65

• fix bug:makes transform ignore filtering more flexible (#274)

  Show description for 0b4c6a5
  

  gavingavin99

  and

  gavinlee

  authored
  0b4c6a5

  Copy full SHA for 0b4c6a5

• support image size control by max_pixels/min_pixels for eagle3 (#272)

  dawnranger

  authored
  bf884ae

  Copy full SHA for bf884ae

Commits on Mar 24, 2026

• feat(spinquant): refactor SpinQuant with CPU offload, parallel fuse, meta device support, and vLLM export (#269)

  gavingavin99

  authored
  99efb57

  Copy full SHA for 99efb57

• support qwen3.5 & transformers>=5.2 (#271)

  linchuanxie

  authored
  e138f05

  Copy full SHA for e138f05

• feat: add QAT pipline (#265)

  ali-88123

  authored
  028f7ab

  Copy full SHA for 028f7ab

Commits on Mar 23, 2026

• [feat]: Support DAQ. (#268)

  Show description for 02b842b
  

  C-KRSW

  and

  krizaltang

  authored
  02b842b

  Copy full SHA for 02b842b

• speculative: add qwen3-vl-8b-eagle3-mrope config (#270)

  dawnranger

  authored
  3bd41ba

  Copy full SHA for 3bd41ba

Commits on Mar 20, 2026

• add int8_dynamic quant config of Qwen VL models (#266)

  irisliu10

  authored
  4fe1230

  Copy full SHA for 4fe1230

Commits on Mar 19, 2026

• support eagle3 offline training with per_device_train_batch_size>1 (#264)

  dawnranger

  authored
  676ae7d

  Copy full SHA for 676ae7d

Commits on Mar 18, 2026

• feat: add SpinQuant offline rotation and integrate with PTQ pipeline (#262)

  gavingavin99

  authored
  26567b2

  Copy full SHA for 26567b2

Commits on Mar 16, 2026

• dflash (#259)

  Show description for 6c91c69
  

  zhzihao

  and

  jiebinzhang

  authored
  6c91c69

  Copy full SHA for 6c91c69

Commits on Mar 12, 2026

• fix pyarrow offset overflow problem (#254)

  dawnranger

  authored
  d3199fd

  Copy full SHA for d3199fd

Commits on Mar 11, 2026

• update script for vllm quant flow (#256)

  liusong1222

  authored
  24644ec

  Copy full SHA for 24644ec

• feat: add end-to-end vLLM W4A8+FP8 mixed quantization pipeline (#255)

  liusong1222

  authored
  9985e28

  Copy full SHA for 9985e28

Commits on Mar 6, 2026

• feat: support calibration with vLLM backend (#251)

  liusong1222

  authored
  10101f4

  Copy full SHA for 10101f4

Commits on Mar 3, 2026

• Update device set of inputs to adapt to text and multimodal calibration (#249)

  irisliu10

  authored
  26a248c

  Copy full SHA for 26a248c

Commits on Feb 28, 2026

• add generate_vlm_hidden_for_draft_model_batch.sh for vlm (#247)

  irisliu10

  authored
  0f7df68

  Copy full SHA for 0f7df68

Commits on Feb 26, 2026

• update TechnicalReport arxiv link (#245)

  Show description for 0e0a6e8
  

  WOODchen7

  and

  woodchenwu

  authored
  0e0a6e8

  Copy full SHA for 0e0a6e8

• update readme (#244)

  yghstill

  authored
  b2f1e8c

  Copy full SHA for b2f1e8c

Commits on Feb 25, 2026

• update readme and code format (#243)

  yghstill

  authored
  3bbee0a

  Copy full SHA for 3bbee0a

Commits on Feb 24, 2026

• feat: Add Windows support with triton-windows and PyTorch fallback (#237)

  Show description for 1b717a4
  

  tori29umai0123

  and

  claude

  authored
  1b717a4

  Copy full SHA for 1b717a4

Commits on Feb 10, 2026

• Update latest news section in README.md (#240)

  WOODchen7

  authored
  ef0c528

  Copy full SHA for ef0c528

Commits on Feb 2, 2026

• fix vlm dataset processor with no_padding formate (#224)

  irisliu10

  authored
  8e66c0e

  Copy full SHA for 8e66c0e

Commits on Jan 29, 2026

• update requirments (#234)

  linchuanxie

  authored
  a044e89

  Copy full SHA for a044e89

Commits on Jan 28, 2026

• [Kunlunxin]support DS R1/V3 w4a8int8 per-channel quantization (#226)

  Show description for 1ea6e3d
  

  lishaoao

  and

  leolishaohao

  authored
  1ea6e3d

  Copy full SHA for 1ea6e3d

Commits on Jan 23, 2026

• support ptpc-int8 (#225)

  linchuanxie

  authored
  ea8540b

  Copy full SHA for ea8540b

Commits on Jan 21, 2026

• fix: initialize add_inputs_embeds to avoid UnboundLocalError in eagle3 (#221)

  liusong1222

  authored
  4e6a606

  Copy full SHA for 4e6a606

Commits on Jan 20, 2026

• support qwen2.5 eagle3 training (#220)

  liusong1222

  authored
  5c9955e

  Copy full SHA for 5c9955e

Commits on Jan 16, 2026

• fix quant type of Qwen3_VL_MoE compressed-tensor to fp8 (#217)

  irisliu10

  authored
  57cb107

  Copy full SHA for 57cb107