Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Aug 22, 2024

• Fix affine quantized tensor to device calls (pytorch#726)

  Show description for 99644e9
  jerryzh168

  authored
  99644e9

  Copy full SHA for 99644e9

• cast local_scale_tensor to fp32 for precompute of fp8 dynamic scaling (pytorch#713)

  Show description for cdb25a4
  

  crcrpar

  and

  msaroufim

  authored
  cdb25a4

  Copy full SHA for cdb25a4

Commits on Aug 21, 2024

• torchao-nightly -> --pre torchao in README (pytorch#723)

  msaroufim

  authored
  5c0e060

  Copy full SHA for 5c0e060

• Enable to(device=device_name) for Uintx (pytorch#722)

  Show description for 9cc81e6
  yiliu30

  authored
  9cc81e6

  Copy full SHA for 9cc81e6

• Update licenses in torchao/experimental

  Show description for d683ba3
  metascroy

  authored
  d683ba3

  Copy full SHA for d683ba3

Commits on Aug 20, 2024

• Bug fixes

  Show description for 227d4bf
  metascroy

  authored
  227d4bf

  Copy full SHA for 227d4bf

• Refactor QAT to use tensor subclasses (pytorch#585)

  Show description for 2c8e3f3
  andrewor14

  authored
  2c8e3f3

  Copy full SHA for 2c8e3f3

• Add high-level operator interface

  Show description for 1909171
  metascroy

  authored
  1909171

  Copy full SHA for 1909171

• added gpu benchmarking script (pytorch#192)

  Show description for 0991ba9
  jcaip

  authored
  0991ba9

  Copy full SHA for 0991ba9

Commits on Aug 19, 2024

• remove outdated TODO from config.py (pytorch#703)

  crcrpar

  authored
  ac8ce4c

  Copy full SHA for ac8ce4c

• Add option to move param to device before quantization (pytorch#699)

  Show description for 477ddb6
  gau-nernst

  authored
  477ddb6

  Copy full SHA for 477ddb6

Commits on Aug 16, 2024

• Relax QAT dtype assertion (pytorch#692)

  Show description for b523f9f
  andrewor14

  authored
  b523f9f

  Copy full SHA for b523f9f

• Refactor superblock code + add final benchmark / eval scripts (pytorch#691)

  Show description for f2c908b
  jcaip

  authored
  f2c908b

  Copy full SHA for f2c908b

• Remove torchao_nightly package (pytorch#661)

  Show description for 4082008
  atalman

  authored
  4082008

  Copy full SHA for 4082008

• Add experimental INT8 quantized training (pytorch#644)

  Show description for 8c6b4f9
  gau-nernst

  authored
  8c6b4f9

  Copy full SHA for 8c6b4f9

• [FEAT] Perf Profiler Update (pytorch#690)

  Show description for 0b66ff0
  jeromeku

  authored
  0b66ff0

  Copy full SHA for 0b66ff0

• Update README.md (pytorch#649)

  jerryzh168

  authored
  ed4c405

  Copy full SHA for ed4c405

Commits on Aug 15, 2024

• Update micro benchmarking code for AQT (pytorch#673)

  Show description for 7f0621d
  jerryzh168

  authored
  7f0621d

  Copy full SHA for 7f0621d

• Add HQQ support (pytorch#605)

  Show description for 18e38f1
  mobicham

  authored
  18e38f1

  Copy full SHA for 18e38f1

• add roofline estimation of float8 gemm + overhead (pytorch#668)

  Show description for 027bf39
  vkuzo

  authored
  027bf39

  Copy full SHA for 027bf39

• Add PyTorch 2.4 tests in CI (pytorch#654)

  msaroufim

  authored
  ffa88a4

  Copy full SHA for ffa88a4

• Fix source version check (pytorch#684)

  msaroufim

  authored
  0b0192e

  Copy full SHA for 0b0192e

• Add BSR subclass +torch.compile and clean up superblock (pytorch#680)

  Show description for b16f0dc
  jcaip

  authored
  b16f0dc

  Copy full SHA for b16f0dc

• Move developer guide file to a folder (pytorch#681)

  Show description for 5998389
  jerryzh168

  authored
  5998389

  Copy full SHA for 5998389

Commits on Aug 14, 2024

• retry version guard fix (pytorch#679)

  Show description for 1acd710
  msaroufim

  authored
  1acd710

  Copy full SHA for 1acd710

• Refactor _quantized_linear for better extensibility (pytorch#634)

  Show description for 582b6d4
  jerryzh168

  authored
  582b6d4

  Copy full SHA for 582b6d4

• Add AffineQuantizedObserver (pytorch#650)

  Show description for 6199f89
  jerryzh168

  authored
  6199f89

  Copy full SHA for 6199f89

• Remove numpy as bitpack dependency (pytorch#677)

  Show description for acfb0da
  msaroufim

  authored
  acfb0da

  Copy full SHA for acfb0da

• Add layout option to woq int4 api (pytorch#670)

  Show description for 009f55f
  Diogo-V

  authored
  009f55f

  Copy full SHA for 009f55f

Commits on Aug 13, 2024

• float8 gemm benchmarks: add option for gpu time (pytorch#666)

  Show description for 174e630
  vkuzo

  authored
  174e630

  Copy full SHA for 174e630

• Spelling fixes (pytorch#662)

  Show description for 88a263a
  jainapurva

  authored
  88a263a

  Copy full SHA for 88a263a

• Move Uintx out of prototype for future extension (pytorch#635)

  Show description for e7fc0ed
  jerryzh168

  authored
  e7fc0ed

  Copy full SHA for e7fc0ed

Commits on Aug 12, 2024

• Add superblock to sparse/prototype (pytorch#660)

  Show description for accbdba
  jcaip

  authored
  accbdba

  Copy full SHA for accbdba

• fix bug with float8 + inference_mode (pytorch#659)

  Show description for d3b8d43
  vkuzo

  authored
  d3b8d43

  Copy full SHA for d3b8d43

• float8 profiling: save torch_logs, and attach them to profiling trace (pytorch#645)

  Show description for 9eef1ae
  vkuzo

  authored
  9eef1ae

  Copy full SHA for 9eef1ae

Pagination

Previous

Next