Code

GaLore and fused kernel prototypes#95

Code

Merged

msaroufim merged 28 commits intopytorch:mainpytorch/ao:mainfrom jeromeku:galore_fusedjeromeku/ao:galore_fusedCopy head branch name to clipboardApr 16, 2024

+3,918

Commits

Commits on Mar 28, 2024

• initial commit

  jeromeku

  committed
  89c707c

  Copy full SHA for 89c707c

• add placeholders for cutlass and triton

  jeromeku

  committed
  d7d19f1

  Copy full SHA for d7d19f1

• update readme

  jeromeku

  committed
  cad0b1b

  Copy full SHA for cad0b1b

• fix versions

  jeromeku

  committed
  0219d87

  Copy full SHA for 0219d87

• minor text edits

  jeromeku

  committed
  de55ffa

  Copy full SHA for de55ffa

Commits on Apr 3, 2024

• clean up

  jeromeku

  committed
  d684670

  Copy full SHA for d684670

• add triton bnb quant kernel and test

  jeromeku

  committed
  fcb0648

  Copy full SHA for fcb0648

• add notes on triton quant kernel

  jeromeku

  committed
  9aa5649

  Copy full SHA for 9aa5649

• refactor code structure

  jeromeku

  committed
  00864f0

  Copy full SHA for 00864f0

• add galore downproj test

  jeromeku

  committed
  f5fae48

  Copy full SHA for f5fae48

• refactor test utils

  jeromeku

  committed
  c3d107b

  Copy full SHA for c3d107b

• add fused kernel tests

  jeromeku

  committed
  436e2b6

  Copy full SHA for 436e2b6

• add fused benchmark

  jeromeku

  committed
  02b52a5

  Copy full SHA for 02b52a5

Commits on Apr 4, 2024

• add dequant kernel

  jeromeku

  committed
  366d9ef

  Copy full SHA for 366d9ef

• update docs

  jeromeku

  committed
  6b39089

  Copy full SHA for 6b39089

Commits on Apr 10, 2024

• add galore memory test

  jeromeku

  committed
  45191aa

  Copy full SHA for 45191aa

Commits on Apr 11, 2024

• add adamw8bit

  jeromeku

  committed
  23c1f65

  Copy full SHA for 23c1f65

• fix README

  jeromeku

  committed
  bf9ee81

  Copy full SHA for bf9ee81

• clean up binaries

  jeromeku

  committed
  7a0994f

  Copy full SHA for 7a0994f

• remove notebook, add instructions to README

  jeromeku

  committed
  25c3875

  Copy full SHA for 25c3875

• remove sample data

  jeromeku

  committed
  22161b7

  Copy full SHA for 22161b7

• Update galore tests

  Show description for ace710d
  jeromeku

  committed
  ace710d

  Copy full SHA for ace710d

• rename galore docs

  jeromeku

  committed
  de040e7

  Copy full SHA for de040e7

Commits on Apr 12, 2024

• More test edits

  Show description for 26ec77d
  jeromeku

  committed
  26ec77d

  Copy full SHA for 26ec77d

• decrease fused matmul parametrizations

  jeromeku

  committed
  adf5250

  Copy full SHA for adf5250

• remove long-running tests

  jeromeku

  committed
  0a86547

  Copy full SHA for 0a86547

• remove tf32 test for now

  jeromeku

  committed
  4c8417e

  Copy full SHA for 4c8417e

Commits on Apr 16, 2024

• Merge branch 'main' into galore_fused

  msaroufim

  authored
  3e7dd86

  Copy full SHA for 3e7dd86