strisunshinewentingwang/pytorch

Pull Requests Commits

Remove call into python API without GIL being held

albanD committed 2 years ago

80eab02a

[easy][PTE] Reduce unnecessary ref count bumps in callstack debug (#72547)

pavithranrao committed 2 years ago

d79aec91

Improve numerical stability of `torch.distributions.wishart.Wishart` (#72059)

nonconvexopt committed 2 years ago

5343cfe9

[DataPipe] Improve .pyi generation (#72829)

NivekT committed 2 years ago

87975d89

[AutoAccept][Codemod][FBSourceClangFormatLinter] Daily `arc lint --take CLANGFORMAT`

CodemodService FBSourceClangFormatLinterBot committed 2 years ago

f395a75c

Revert D34111109: [FSDP] Implement apply()

malfet committed 2 years ago

ccdff4c4

Set `BLAS_LIBRARIES` to `${MKL_LIBRARIES}` for MKL case (#72806)

atalman committed 2 years ago

Verified 17b3ba14

[Join][BE] Fix typo; remove obsolete method (#72886)

awgu committed 2 years ago

59dd84ca

[FSDP] Implement apply() (#72600)

rohan-varma committed 2 years ago

1f29b313

Port `amax` to structured kernel (#72124)

Ansley Ussery committed 2 years ago

e4214929

[Perf] Reduce unnecessary ref count bumps (#72523)

xta0 committed 2 years ago

85d7e73a

Revert D34106940: [ZeRO] Add ctor support for multiple param groups

malfet committed 2 years ago

84cb810b

(2/2) Make TorchScript Preserve Fully Qualified Class Name for Python Exceptions: frontend change (#72899)

shunting314 committed 2 years ago

763ad1bf

Make fusion strategy api public (#72639)

Elias Ellison committed 2 years ago

f8a2efc1

Move dyn fusion api to jit/api/module/ (#72638)

Elias Ellison committed 2 years ago

1750c017

[Quant] Add qint32 quantization support (#72472)

terrychenism committed 2 years ago

f67cf035

[Qunat] Add ConvTranspose reference module (#72473)

terrychenism committed 2 years ago

7a031ec1

[Quant][core][devs] Separated implementations for quantized & non-quantized tensors in reflection_pad2d_cpu (#72442)

dzdang committed 2 years ago

ad38b92f

[monitored barrier] Slight logging enhancement (#72754)

rohan-varma committed 2 years ago

ec3a5ca6

[Checkpoint] Rename file (#72748)

rohan-varma committed 2 years ago

aeacf910

[FSDP] Improved shape unflattening test (#72573)

rohan-varma committed 2 years ago

08889b24

[FSDP] Fix summon_full_params when not sharded (#72572)

rohan-varma committed 2 years ago

b01d1ad1

Rename `Typed/UntypedStorage` to `_Typed/_UntypedStorage` (#72540)

kurtamohler committed 2 years ago

8e7fe876

Fix the doc of PostLocalSGDState (#72792)

wayi1 committed 2 years ago

8b084781

[Kineto][Bug Fix] Avoid picking up old CUPTI headers (#72761)

aaronenyeshi committed 2 years ago

8a43aa95

Fix vjpvmap for linalg.svdvals (#72811)

soulitzer committed 2 years ago

277c4c9d

Remove xfail for trapz and trapezoid on meta device (#72677)

soulitzer committed 2 years ago

67adc0cb

be: Remove unused docker folder (#72884)

seemethere committed 2 years ago

93a8bbbc

[quant][fx][improvement] Add lowering support for BatchNormQuantizeHandler (#72490)

jerryzh168 committed 2 years ago

3d377fb4

`linalg_det_singular`: modify samples such that CUDA IMA dissapears. (#72585)

nikitaved committed 2 years ago

961bbe1c

Older