Commits

Branch selector

 mpk

User selector

All users

Datepicker

All time

Commit History

Commits on Nov 12, 2025

• Merge pull request #575 from xiaozr/mpk

  Show description for ef36cb4
  xiaozr

  authored
  ef36cb4

  Copy full SHA for ef36cb4

• rvt page_indices load

  xiaozr

  committed
  efacfa8

  Copy full SHA for efacfa8

• Fix misalignment issue caused by attention kernel (#572)

  Show description for 16dcdfb
  xinhaoc

  authored
  16dcdfb

  Copy full SHA for 16dcdfb

Commits on Nov 11, 2025

• fix compilation err (#570)

  NorthmanPKU

  authored
  0228d17

  Copy full SHA for 0228d17

• [MPK] Fix rotary_embedding and llama demo (#564)

  Show description for 7541607
  xiaozr

  authored
  7541607

  Copy full SHA for 7541607

• [MPK] Qwen3-30B-A3B MoE demo (#563)

  Show description for 62239a8
  

  18 peopleauthored
  62239a8

  Copy full SHA for 62239a8

• Update Ampere attention kernel to save shared memory usage (#553)

  Show description for 5478a65
  xinhaoc

  authored
  5478a65

  Copy full SHA for 5478a65

Commits on Nov 10, 2025

• blackwell bug fix regarding try_wait_barrier (#567)

  Show description for 7514c44
  

  JackFram

  and

  Zhihao Zhang

  authored
  7514c44

  Copy full SHA for 7514c44

Commits on Nov 8, 2025

• add MIRAGE_BACKEND_USE_CUDA flag to kernel compilation (#565)

  dtunai

  authored
  b3f1076

  Copy full SHA for b3f1076

Commits on Nov 4, 2025

• fix profiler and typos (#562)

  yzhou442

  authored
  3738646

  Copy full SHA for 3738646

Commits on Nov 2, 2025

• Refator linear kernel (#548)

  Show description for 856a10c
  wenqiny

  authored
  856a10c

  Copy full SHA for 856a10c

Commits on Oct 30, 2025

• [MPK] Hopper Linear Optimization (#561)

  Show description for b22a9bc
  yzhou442

  authored
  b22a9bc

  Copy full SHA for b22a9bc

Commits on Oct 29, 2025

• Fix missing use_cutlass_kernel arg in llama3 demo (#559)

  byungsoo-oh

  authored
  e15b0e1

  Copy full SHA for e15b0e1

• [MPK] Update schedulers to allow calling persistent kernel multiple times (#557)

  Show description for 5e75bb4
  jiazhihao

  authored
  5e75bb4

  Copy full SHA for 5e75bb4

Commits on Oct 27, 2025

• Update Hopper Group GEMM (#551)

  Show description for 1c228ea
  yzhou442

  authored
  1c228ea

  Copy full SHA for 1c228ea

Commits on Oct 26, 2025

• [MPK] Update tests/runtime_python to support unit tests for Ampere GPUs (#550)

  Show description for c0d55d3
  jiazhihao

  authored
  c0d55d3

  Copy full SHA for c0d55d3

• fix: fix tp > 1 model convert when tie_word_embeddings is True (#549)

  Show description for 8198a56
  izhuhaoran

  authored
  8198a56

  Copy full SHA for 8198a56

Commits on Oct 25, 2025

• MoE Hopper Group GEMM Implementation (#547)

  Show description for 87e3aa1
  yzhou442

  authored
  87e3aa1

  Copy full SHA for 87e3aa1

• Blackwell MoE Task Implementation and unit tests update (#546)

  Show description for d3b1fbb
  

  18 peopleauthored
  d3b1fbb

  Copy full SHA for d3b1fbb

Commits on Oct 22, 2025

• feat: support more HIDDEN_DIM for more qwen3 model (#544)

  Show description for 7e4f9e7
  izhuhaoran

  authored
  7e4f9e7

  Copy full SHA for 7e4f9e7

• refactor swizzle, it reduce ~10% inst and improve >5% performance for linear (#541)

  wenqiny

  authored
  9ba6947

  Copy full SHA for 9ba6947

Commits on Oct 20, 2025

• [MPK] Add Llama3 family demo (#535)

  Show description for e36b734
  xiaozr

  authored
  e36b734

  Copy full SHA for e36b734

• fix: fix demo hopper miss use_cutlass_kernel arg (#539)

  Show description for 0c1f88c
  izhuhaoran

  authored
  0c1f88c

  Copy full SHA for 0c1f88c

Commits on Oct 16, 2025

• Optimize multitoken_paged_attention_hopper.cuh && linear_swapAB_hopper.cuh (#536)

  Show description for 6cdb8f3
  yzhou442

  authored
  6cdb8f3

  Copy full SHA for 6cdb8f3

• fix a bug for mpk ptx linear kernel (#533)

  Show description for b8d7213
  

  wenqiny

  and

  jiazhihao

  authored
  b8d7213

  Copy full SHA for b8d7213

Commits on Oct 15, 2025

• feat: support ignore eos for qwen3 demo (#530)

  Show description for 2746c21
  izhuhaoran

  authored
  2746c21

  Copy full SHA for 2746c21

• perf: support more new prefill add per round (#532)

  Show description for 7a91cd2
  izhuhaoran

  authored
  7a91cd2

  Copy full SHA for 7a91cd2

Commits on Oct 13, 2025

• [MPK] Fix multi-GPU issues (#527)

  Show description for bb075eb
  jiazhihao

  authored
  bb075eb

  Copy full SHA for bb075eb

Commits on Oct 12, 2025

• Refactor task implementation files to support multiple GPU architectures (#523)

  Show description for a9350c9
  

  jiazhihao

  and

  goliaro

  authored
  a9350c9

  Copy full SHA for a9350c9

Commits on Oct 11, 2025

• MPK SM100 Linear Task added (#514)

  Show description for 5162bfb
  

  16 peopleauthored
  5162bfb

  Copy full SHA for 5162bfb

Commits on Oct 10, 2025

• Add cutlass version Hopper kernels (#497)

  Show description for d32cc9c
  

  xinhaoc

  and

  yzhou442

  authored
  d32cc9c

  Copy full SHA for d32cc9c

Commits on Oct 9, 2025

• upd (#521)

  xinhaoc

  authored
  f171892

  Copy full SHA for f171892

Commits on Oct 8, 2025

• add documentation for quantization v1 (#518)

  Show description for cfae093
  Hengrui-Chen

  authored
  cfae093

  Copy full SHA for cfae093

Commits on Oct 7, 2025

• Preload tasks (#511)

  Show description for 9fd10d5
  

  4 peopleauthored
  9fd10d5

  Copy full SHA for 9fd10d5

Commits on Oct 6, 2025

• Display task graph script (#517)

  anthony-yip

  authored
  cc9c3c2

  Copy full SHA for cc9c3c2

Pagination

Previous

Next