Commits

Branch selector

 7c65d88

User selector

All users

Datepicker

All time

Commit History

Commits on Sep 6, 2024

• [Float8Quant] Add rowwise scaling option to float8 dyanmic quant

  Show description for 7c65d88
  drisspg

  committed
  7c65d88

  Copy full SHA for 7c65d88

• Add sparse marlin AQT layout (#621)

  Show description for 65d86c6
  

  Diogo-V

  and

  jcaip

  authored
  65d86c6

  Copy full SHA for 65d86c6

• [StaticQuant] add a linear observer class and test (#807)

  drisspg

  authored
  422301b

  Copy full SHA for 422301b

• Lowbit custom torch op

  Show description for 038a0a2
  metascroy

  authored
  038a0a2

  Copy full SHA for 038a0a2

Commits on Sep 5, 2024

• Add static quant to float8 (#787)

  Show description for a246d87
  jainapurva

  authored
  a246d87

  Copy full SHA for a246d87

• Update version.txt to 0.6.0 (#817)

  andrewor14

  authored
  740c6b3

  Copy full SHA for 740c6b3

• [StaticQuant] Update how block_size is calculated with Observers (#815)

  drisspg

  authored
  a1b3e67

  Copy full SHA for a1b3e67

• Add uintx quant to generate and eval (#811)

  Show description for e05635e
  jerryzh168

  authored
  e05635e

  Copy full SHA for e05635e

• Update main README.md with more current float8 speedup (#816)

  vkuzo

  authored
  9d5527a

  Copy full SHA for 9d5527a

• [Low-bit optim] Improve compile time + Fix PyTorch 2.3 support for 4-bit optim (#812)

  Show description for 1e7f132
  gau-nernst

  authored
  1e7f132

  Copy full SHA for 1e7f132

• Update README.md (#814)

  Show description for 599319f
  HDCharles

  authored
  599319f

  Copy full SHA for 599319f

• adding kv cache quantization to READMEs (#813)

  Show description for 62c239c
  HDCharles

  authored
  62c239c

  Copy full SHA for 62c239c

• make float8 README.md examples standalone (#809)

  Show description for 1c488e8
  vkuzo

  authored
  1c488e8

  Copy full SHA for 1c488e8

• int4 fixes and improvements (#804)

  Show description for 317392d
  HDCharles

  authored
  317392d

  Copy full SHA for 317392d

Commits on Sep 4, 2024

• [Float8] Make Inference and Training code independent (#808)

  drisspg

  authored
  848e123

  Copy full SHA for 848e123

• make f8 roofline script calculate observed overhead (#734)

  Show description for 9d92434
  vkuzo

  authored
  9d92434

  Copy full SHA for 9d92434

• Add Auto-Round support (#581)

  Show description for f5703b0
  yiliu30

  authored
  f5703b0

  Copy full SHA for f5703b0

• Add more information to quantized linear module and added some logs (#782)

  Show description for 0987dd6
  jerryzh168

  authored
  0987dd6

  Copy full SHA for 0987dd6

Commits on Sep 3, 2024

• Move more utils to TorchAOBaseTensor (#784)

  Show description for 9db1660
  jerryzh168

  authored
  9db1660

  Copy full SHA for 9db1660

• [Lint] Format and Lint nf4 file and test (#800)

  drisspg

  authored
  988f56d

  Copy full SHA for 988f56d

• [NF4] Freeze dataclass | prep for better pt2 support (#799)

  drisspg

  authored
  4d8cbc1

  Copy full SHA for 4d8cbc1

• Add 5bit packing routines

  Show description for 6080986
  metascroy

  authored
  6080986

  Copy full SHA for 6080986

• Add 2bit packing routines

  Show description for d980858
  metascroy

  authored
  d980858

  Copy full SHA for d980858

• Fixed the llama model (#769)

  Show description for 8c18489
  yiliu30

  authored
  8c18489

  Copy full SHA for 8c18489

• Allow quantized linear registration in a different file (#783)

  Show description for e15e509
  jerryzh168

  authored
  e15e509

  Copy full SHA for e15e509

Commits on Sep 2, 2024

• Add performance section to float8 README.md (#794)

  vkuzo

  authored
  e2dad4a

  Copy full SHA for e2dad4a

• Pin PT version: Fix FPX Inductor error (#790)

  Show description for 287458c
  msaroufim

  authored
  287458c

  Copy full SHA for 287458c

• [low-bit optim] Fix Adam4bit support on PyTorch 2.3 and 2.4. Update AdamFp8 torch requirement (#755)

  Show description for 65f660d
  gau-nernst

  authored
  65f660d

  Copy full SHA for 65f660d

• [Low-bit optim] Add Llama2-7B finetune benchmarks (#746)

  Show description for e5246fc
  

  gau-nernst

  and

  msaroufim

  authored
  e5246fc

  Copy full SHA for e5246fc

Commits on Aug 30, 2024

• add in formatting skip everything (#779)

  Show description for ba2d3b1
  drisspg

  authored
  ba2d3b1

  Copy full SHA for ba2d3b1

• Add Float8 Weight Only and FP8 weight + dynamic activation (#740)

  Show description for d0e6246
  drisspg

  authored
  d0e6246

  Copy full SHA for d0e6246

Commits on Aug 29, 2024

• [reland] Refactor quant_llm to work with affine quantized tensor (#696) (#772)

  Show description for 05224a9
  jerryzh168

  authored
  05224a9

  Copy full SHA for 05224a9

• torchao::parallel_for backends

  Show description for 09a5e54
  metascroy

  authored
  09a5e54

  Copy full SHA for 09a5e54

Commits on Aug 28, 2024

• Update method names to support intx and floatx changes (#775)

  jainapurva

  authored
  cfabc13

  Copy full SHA for cfabc13

• add readme and code refactor (#776)

  Show description for f538027
  Hanxian97

  authored
  f538027

  Copy full SHA for f538027

Pagination

Previous

Next