Repository navigation

Commits

History for

onmain

Commits on Oct 8, 2024
- Revert "Rename Layout -> TensorImpl" (#1040)
  Mark Saroufim
  authored
  745085f
- Rename Layout -> TensorImpl (#1028)
  jainapurva
  authored
  cc8bf85
- Optimize 3-bit packing
  metascroy
  authored
  93ff876
Commits on Oct 7, 2024
- float8 training axiswise scaling support with per-gemm-argument configuration (#940)
  vkuzo
  authored
  dec0313
- add axiswise scaling to Float8Linear (#920)
  vkuzo
  authored
  e76db70
- Activation Aware Weight Quantization (AWQ) (#743)
  vayuda
  authored
  f81fe11
- Dynamic Float8 benchmarking llama (#1017)
  jainapurva
  authored
  92dd5f5
- add axiswise granularity to Float8Tensor (#919)
  vkuzo
  authored
  52d27a1
- Refactor `tiktoken` import bare except (#1024)
  Matthew Hoffman
  authored
  e7331ab
- Use `importlib.util.find_spec` to check if `lm_eval` is installed instead of trying to import it (#1023)
  Matthew Hoffman
  authored
  e2301e9
Commits on Oct 6, 2024
- [low-bit optim] Fix load state dict when device is different (#1021)
  gau-nernst
  authored
  c187f87
Commits on Oct 4, 2024
- Add Float8 support for AQT tensor parallel (#1003)
  jainapurva
  authored
  9e2a253
- training ir torchao migration
  yushangdi
  authored
  d2982d5
Commits on Oct 3, 2024
- Introduce 6-bit quantization for Llama in torchchat
  ramreddymounica
  authored
  9ce7ebb
- sync with fbcode (#1005)
  metascroy
  authored
  8945fb3
Commits on Oct 2, 2024
- Add float8_static_activation_float8_weight to user-facing API (#994)
  jainapurva
  authored
  09b8b3c
- Use checkpoint to enforece the recomputation of fp8 weight
  y-sq
  authored
  e919558
- Add example for awq like flow (#990)
  jerryzh168
  authored
  378e6a8
- Float8 dynamic autoquant (#946)
  jainapurva
  authored
  9229df9
Commits on Oct 1, 2024
- Gptq mt refactor (#914)
  
  danielpatrickhug
  and
  HDCharles
  authored
  83d5b63
- Add int4 weight-only embedding QAT (#947)
  andrewor14
  authored
  5a4857e
- [float8] improve eager numerics for dynamic scales and gets on par with torch.compile (#904)
  
  authored
  71be315
- add ruff lint rule to remove unused imports via ruff (#969)
  jimexist
  authored
  1137f39
- Print args/kwargs types for utils and examples (#963)
  jerryzh168
  authored
  20f3796
- Don't call dequantize in `__repr__` (#965)
  jerryzh168
  authored
  b4055da
- BitNet b1.58 training (#930)
  gau-nernst
  authored
  a382752
Commits on Sep 30, 2024
- Change arg order in lowbit linear ops to match aten (#982)
  metascroy
  authored
  ae3e7c6
- Skip test_choose_qparams_token_asym on pt 2.6 (#981)
  Mark Saroufim
  authored
  68e1886
- Add arm neon guards
  metascroy
  authored
  60ffb86
- Change op names to linear_8bit_act_xbit_weight
  metascroy
  authored
  b983f7d
- Add new evaluation metrics (#934)
  jainapurva
  authored
  ae49375
Commits on Sep 29, 2024
- [low-bit optim] Update docs on supported platforms and caveats (#971)
  gau-nernst
  authored
  c0a81f9
- SM75 (Turing) support for FP6 kernel (#942)
  tobiasvanderwerff
  authored
  96e8fee
Commits on Sep 28, 2024
- Fix WOQ int8 failures (#884)
  leslie-fang-intel
  authored
  39ce823
Commits on Sep 27, 2024
- Supporting tensor parallelism for int8 weight only quant (#939)
  
  jerryzh168
  and
  kwen2501
  authored
  72d2518

Breadcrumbs

User selector

Datepicker

Commit history

Commits on Oct 8, 2024

Commits on Oct 7, 2024

Commits on Oct 6, 2024

Commits on Oct 4, 2024

Commits on Oct 3, 2024

Commits on Oct 2, 2024

Commits on Oct 1, 2024

Commits on Sep 30, 2024

Commits on Sep 29, 2024

Commits on Sep 28, 2024

Commits on Sep 27, 2024