Repository navigation

Cpu fused kernel - #1804

Merged

matthewdouglas merged 83 commits into

bitsandbytes-foundation:mainfrom

jiqing-feng:cpu_fused_kernel

Nov 26, 2025

Merged

Cpu fused kernel#1804
matthewdouglas merged 83 commits into
bitsandbytes-foundation:mainfrom
jiqing-feng:cpu_fused_kernel

Commits on Oct 28, 2025
- add template to support more dtypes
  jiqing-feng
  committed
  6be1412
- update cmake list
  jiqing-feng
  committed
  252ac0f
- fix typo
  jiqing-feng
  committed
  f98c9e5
- fix compile cpu
  jiqing-feng
  committed
  902bf35
Commits on Oct 29, 2025
- make different dtype works
  jiqing-feng
  committed
  fef8459
- use bf16 on CPU
  jiqing-feng
  committed
  55cbaa0
- fix state2 dtype
  jiqing-feng
  committed
  bbef95b
Commits on Oct 30, 2025
- remove torch
  jiqing-feng
  committed
  e842513
- rm torch
  jiqing-feng
  committed
  d4473fa
- enable float to bf16
  jiqing-feng
  committed
  dea8dd6
- rm dequantizeBlockwise4bitCpu
  jiqing-feng
  committed
  e9bb4fe
- fix check
  jiqing-feng
  committed
  cdc8d5e
- enable dequant 4bit kernel
  jiqing-feng
  committed
  baacfac
- fix typo
  jiqing-feng
  committed
  eec3521
- fix typo
  jiqing-feng
  committed
  d7cc1c5
- fix dequantize
  jiqing-feng
  committed
  124b754
- fix
  jiqing-feng
  committed
  0f918c7
- fix
  jiqing-feng
  committed
  e1a8b20
- test
  jiqing-feng
  committed
  eab45c8
- fix
  jiqing-feng
  committed
  d9f5dd8
- fix
  jiqing-feng
  committed
  070f8a0
- fix
  jiqing-feng
  committed
  a84addf
- fix
  jiqing-feng
  committed
  c4bb660
- fix
  jiqing-feng
  committed
  4ba13fd
Commits on Oct 31, 2025
- change input param
  jiqing-feng
  committed
  c0d05ec
- fix typo
  jiqing-feng
  committed
  62a16a6
- fix input param
  jiqing-feng
  committed
  d9ad828
- spliut 8bit and 4bit
  jiqing-feng
  committed
  09ed6cb
- fix typo
  jiqing-feng
  committed
  a3f7b61
- fix typo
  jiqing-feng
  committed
  4708470
- fix input params
  jiqing-feng
  committed
  1dfe9f7
- fix input params
  jiqing-feng
  committed
  00289c4
- fix
  jiqing-feng
  committed
  a2578ba
- fix typo
  jiqing-feng
  committed
  72033dc
- enable dequant4bit
  jiqing-feng
  committed
  1c20ae8
- fix
  jiqing-feng
  committed
  7552fe2
- fix
  jiqing-feng
  committed
  8b32a39
- fix reverse
  jiqing-feng
  committed
  8f1cc36
Commits on Nov 3, 2025
- fix dequant 4bit fallback path
  jiqing-feng
  committed
  49d242a
- fix fp4 dequant
  jiqing-feng
  committed
  4a9a6dc
Commits on Nov 4, 2025
- Merge branch 'main' into cpu_kernel
  jiqing-feng
  authored
  6bcd19e
Commits on Nov 5, 2025
- rm _Float16
  jiqing-feng
  committed
  d7e981d
Commits on Nov 6, 2025
- tmp codes
  jiqing-feng
  committed
  48739b0
Commits on Nov 7, 2025
- enable gemv
  jiqing-feng
  committed
  f784be8
- change to 4bit dequant
  jiqing-feng
  committed
  92192c9
- fix def
  jiqing-feng
  committed
  bd02e71
- fix type
  jiqing-feng
  committed
  8520069
- fix absmax dtype
  jiqing-feng
  committed
  e921cbb
- fix type
  jiqing-feng
  committed
  9b5d97a
- fix compile and type
  jiqing-feng
  committed
  fd6cff1
- enable gemv
  jiqing-feng
  committed
  46d6e47
- fix shape
  jiqing-feng
  committed
  3271c30
- fix lib name
  jiqing-feng
  committed
  176a2b6
- debug
  jiqing-feng
  committed
  196984a
Commits on Nov 11, 2025
- update
  jiqing-feng
  committed
  7652115
Commits on Nov 12, 2025
- enable gemv 4bit bf16
  jiqing-feng
  committed
  ea0e649
Commits on Nov 13, 2025
- enable avx512 check
  jiqing-feng
  committed
  9277d24
- fix check
  jiqing-feng
  committed
  4fb315b
- fix endif
  jiqing-feng
  committed
  81f1984
- fix format
  jiqing-feng
  committed
  0f78bad
- fix format
  jiqing-feng
  committed
  fcb8456
- fix def
  jiqing-feng
  committed
  c5e1894
Commits on Nov 14, 2025
- rebase
  jiqing-feng
  committed
  f2029c6
- fix position
  jiqing-feng
  committed
  df1d669
- fix format
  jiqing-feng
  committed
  bb3ac8d
- rm duplicated func
  jiqing-feng
  committed
  26b5685
Commits on Nov 17, 2025
- Merge branch 'main' into cpu_fused_kernel
  jiqing-feng
  authored
  445725b
- rm useless code comments
  jiqing-feng
  committed
  580010c
Commits on Nov 19, 2025
- fix out shape
  jiqing-feng
  committed
  57b89bf
- Merge branch 'main' into cpu_fused_kernel
  jiqing-feng
  authored
  302a5fe
Commits on Nov 20, 2025
- fix comments
  jiqing-feng
  committed
  de5fb9c
- add reverse format
  jiqing-feng
  committed
  6858a90
- check avx512bf15
  jiqing-feng
  committed
  3b3d609
- fix has_avx512bf16
  jiqing-feng
  committed
  fbb911b
- fix tests
  jiqing-feng
  committed
  3179b42
- fix absmax shhape
  jiqing-feng
  committed
  0c88d43
- fix compile
  jiqing-feng
  committed
  feb8ad2
- fix tests
  jiqing-feng
  committed
  c6b714d
- fix test_gemv
  jiqing-feng
  committed
  5497111
Commits on Nov 24, 2025
- Merge branch 'main' into cpu_fused_kernel
  jiqing-feng
  authored
  0045c4b
Commits on Nov 25, 2025
- disable binsearch
  jiqing-feng
  committed
  bdb25c0
- fix lint
  jiqing-feng
  committed
  6cec12d
Commits on Nov 26, 2025
- fix save
  jiqing-feng
  committed
  692a8e1

Commits

Commits on Oct 28, 2025

Commits on Oct 29, 2025

Commits on Oct 30, 2025

Commits on Oct 31, 2025

Commits on Nov 3, 2025

Commits on Nov 4, 2025

Commits on Nov 5, 2025

Commits on Nov 6, 2025

Commits on Nov 7, 2025

Commits on Nov 11, 2025

Commits on Nov 12, 2025

Commits on Nov 13, 2025

Commits on Nov 14, 2025

Commits on Nov 17, 2025

Commits on Nov 19, 2025

Commits on Nov 20, 2025

Commits on Nov 24, 2025

Commits on Nov 25, 2025

Commits on Nov 26, 2025