Commits

Branch selector

 main

User selector

tdoublep

Datepicker

All time

Commit History

Commits on Nov 12, 2025

• [Model] [Config] Correctly identify granite-4.0-micro as non-hybrid model (#28563)

  Show description for 64d57c3
  tdoublep

  authored
  64d57c3

  Copy full SHA for 64d57c3

Commits on Nov 3, 2025

• [Hybrid] Pass kernel block size to builders (#27753)

  Show description for 18961c5
  tdoublep

  authored
  18961c5

  Copy full SHA for 18961c5

Commits on Oct 7, 2025

• Revert #24446 and #26168 (#26332)

  Show description for 31a4b3e
  tdoublep

  authored
  31a4b3e

  Copy full SHA for 31a4b3e

Commits on Oct 6, 2025

• [V1] [Hybrid] Some additional clean-up in Mamba2 prefix caching (#26222)

  Show description for 778f554
  tdoublep

  authored
  778f554

  Copy full SHA for 778f554

• [CI] Add comment about the single cudagraph capture size that is used (#26252)

  tdoublep

  authored
  d3c8429

  Copy full SHA for d3c8429

Commits on Oct 4, 2025

• [V1] [Hybrid] Remove code to override default CUDA graph configuration (#26226)

  Show description for ed3aeb2
  tdoublep

  authored
  ed3aeb2

  Copy full SHA for ed3aeb2

Commits on Oct 3, 2025

• [CI] Fix distributed hybrid tests in CI (#26155)

  Show description for 0e93ac0
  tdoublep

  authored
  0e93ac0

  Copy full SHA for 0e93ac0

Commits on Oct 2, 2025

• Change size of single CUDA graph for CI to 4 (#26089)

  Show description for be8921f
  tdoublep

  authored
  be8921f

  Copy full SHA for be8921f

Commits on Sep 29, 2025

• [Kernel] Chunk-aligned mamba2 (#24683)

  tdoublep

  authored
  fea3e47

  Copy full SHA for fea3e47

• Add Phi4FlashForCausalLM to _PREVIOUSLY_SUPPORTED_MODELS (#25832)

  Show description for 219cfbe
  tdoublep

  authored
  219cfbe

  Copy full SHA for 219cfbe

Commits on Sep 23, 2025

• [Kernel] [Mamba] Remove BLOCK_H=1 from list of tuneable configurations for `_chunk_cumsum_fwd_kernel` (#25197)

  Show description for 5e25b12
  

  tdoublep

  and

  cyang49

  authored
  5e25b12

  Copy full SHA for 5e25b12

• [V0 Deprecation] Remove placeholder attn (#25510)

  Show description for 969b4da
  tdoublep

  authored
  969b4da

  Copy full SHA for 969b4da

• [V1] Remove V0 code paths for Hybrid models (#25400)

  Show description for a903669
  tdoublep

  authored
  a903669

  Copy full SHA for a903669

Commits on Sep 2, 2025

• [CI] Enable all hf transformers baselines in test_hybrid (#23936)

  Show description for d328f78
  tdoublep

  authored
  d328f78

  Copy full SHA for d328f78

Commits on Aug 30, 2025

• [V1] [Hybrid] Move MiniMaxLinearAttention into layers/mamba (#23831)

  Show description for 4071c76
  

  tdoublep

  and

  DarkLight1337

  authored
  4071c76

  Copy full SHA for 4071c76

Commits on Aug 29, 2025

• [Docs] [V1] [Hybrid] Add new documentation re: contributing mamba-based models (#23824)

  Show description for 1c26b42
  tdoublep

  authored
  1c26b42

  Copy full SHA for 1c26b42

Commits on Aug 28, 2025

• [CI] Fix linting error on main (#23835)

  Show description for 1f096f9
  tdoublep

  authored
  1f096f9

  Copy full SHA for 1f096f9

Commits on Aug 27, 2025

• [V1] [Hybrid] Enable compile and piecewise CUDA graph for MiniMax-Text models (#22589)

  Show description for dd58932
  tdoublep

  authored
  dd58932

  Copy full SHA for dd58932

• [V1] [Hybrid] Disable prefix caching by default for hybrid or mamba-based models (#23716)

  Show description for 704432a
  tdoublep

  authored
  704432a

  Copy full SHA for 704432a

Commits on Aug 26, 2025

• [V1] [Hybrid] Enable Full CUDA graph by default for hybrid models in V1 (#22594)

  Show description for 5f1af97
  tdoublep

  authored
  5f1af97

  Copy full SHA for 5f1af97

• [Docs] [V1] [Hybrid] Update docs to remove FlashInfer constraint for hybrid models (#23665)

  Show description for 227e231
  tdoublep

  authored
  227e231

  Copy full SHA for 227e231

Commits on Aug 19, 2025

• [Misc] Add @tdoublep as a maintainer of hybrid model and Triton-attention related code (#23122)

  Show description for 95e3095
  tdoublep

  authored
  95e3095

  Copy full SHA for 95e3095

Commits on Aug 15, 2025

• [BugFix] Fix regression caused by mamba state dtype PR (#22998)

  Show description for f5d412b
  tdoublep

  authored
  f5d412b

  Copy full SHA for f5d412b

• [V1] [Hybrid] Support using float32 for state in Hybrid Models (Mamba2, Mamba1, Minimax) (#22928)

  Show description for 75531a6
  

  4 peopleauthored
  75531a6

  Copy full SHA for 75531a6

Commits on Aug 14, 2025

• [CI] [Hybrid] Bump min transformers version for Bamba and Jamba (#22908)

  Show description for ab9f2cf
  tdoublep

  authored
  ab9f2cf

  Copy full SHA for ab9f2cf

Commits on Aug 10, 2025

• [V1] [Hybrid] Enable Full CUDA Graph (decode-only) for Mamba layers (#21401)

  Show description for 61f67d8
  tdoublep

  authored
  61f67d8

  Copy full SHA for 61f67d8

Commits on Aug 9, 2025

• [CI] [Hybrid] Speed up hybrid models test by removing large models (#22563)

  Show description for 1bf5e1f
  tdoublep

  authored
  1bf5e1f

  Copy full SHA for 1bf5e1f

• Update docs for Minimax-Text support (#22562)

  Show description for 2be07a0
  tdoublep

  authored
  2be07a0

  Copy full SHA for 2be07a0

• [V1] [Hybrid] Support Minimax-Text-01 in V1 (#22151)

  Show description for 6ade99e
  tdoublep

  authored
  6ade99e

  Copy full SHA for 6ade99e

• Remove mamba_ssm from vLLM requirements; install inside test container using `--no-build-isolation` (#22541)

  Show description for 8a0ffd6
  tdoublep

  authored
  8a0ffd6

  Copy full SHA for 8a0ffd6

Commits on Aug 8, 2025

• [Bugfix] Update FA commit hash (#22546)

  Show description for bd875d2
  tdoublep

  authored
  bd875d2

  Copy full SHA for bd875d2

Commits on Aug 2, 2025

• [V1] [Hybrid] Validate compatibility of attention backend batch reordering at init time (#21557)

  Show description for 4abfd87
  tdoublep

  authored
  4abfd87

  Copy full SHA for 4abfd87

Commits on Jul 22, 2025

• [V1] [Hybrid] Add new test to verify that hybrid views into KVCacheTensor are compatible (#21300)

  Show description for 488d8a9
  tdoublep

  authored
  488d8a9

  Copy full SHA for 488d8a9

Commits on Jul 19, 2025

• [Docs] [V1] Update docs to remove enforce_eager limitation for hybrid models. (#21233)

  Show description for 2b504eb
  tdoublep

  authored
  2b504eb

  Copy full SHA for 2b504eb

• [V1] [Hybrid] Enable piecewise CUDA Graph for mamba layers (#21194)

  Show description for 881e3cb
  tdoublep

  authored
  881e3cb

  Copy full SHA for 881e3cb

Pagination

Previous

Next