strisunshinewentingwang/pytorch

Pull Requests Commits

Ignore .numpy() under FakeTensorMode() (#120261)

Thiago Crepaldi committed 212 days ago

3588e7f2

[BE][EZ] Flatten preprocessor hierarchy (#120422)

malfet committed 212 days ago

f9eb66e1

[BE][optim] Simplify _init_group. (#120055)

ttsugriy committed 212 days ago

1c7ba330

[DeviceMesh] Ensure mesh tensor is a cpu tensor (#120046)

wz337 committed 212 days ago

5603d953

[ROCm] replace ROCmLoops.cuh with hipified CUDALoops.cuh (#120101)

jeffdaily committed 212 days ago

c11bd724

Use privateuseone during external module register test (#120399)

dilililiwhy committed 212 days ago

77692736

highlight readme code block (#120228)

KimBioInfoStudio committed 212 days ago

edd03f97

[Dynamic] Fix dynamic shape size inspection bug (#120341)

oulgen committed 212 days ago

1eae8950

Temporarily support ranks + tag as pg identifier in native funcol (#120226)

yifuwang committed 212 days ago

11e4a926

Make tests using CommDebugMode work for both legacy and native funcol (#120070)

yifuwang committed 212 days ago

5a3e1957

Prepare test_dtensor.py for native funcol migration (#120043)

yifuwang committed 212 days ago

a4c5f48b

Change the .clone() in native funcol's all_reduce to use at::MemoryFormat::Contiguous (#120042)

yifuwang committed 212 days ago

1c9fc720

[cond] make sure subgraphs in cond are decomposed according to current decomp table (#120366)

ydwu4 committed 212 days ago

7b8f6736

Fix the default value of side in torch.searchsorted (#120066)

lancerts committed 212 days ago

680cfec2

[FSDP2] Removed `super().__setattr__` call (#120340)

awgu committed 212 days ago

c37d07a1

[inductor] decompose memory bound mm (#120047)

jackiexu1992 committed 212 days ago

2ba798df

modify comment of SparseTensor coalesce (#120221)

huihoaan committed 212 days ago

ce807c17

Add code example for torch.stack() (#120304)

onel committed 212 days ago

bb72bfe2

Fix rendering in the doc of `PackedSequence` (#120385)

yuanx749 committed 212 days ago

ca64f7cb

[inductor] improve kernel metadata logging (#120274)

shunting314 committed 212 days ago

a77226aa

[profiler] Add kineto init delay when used in daemon mode (#120276)

briancoutinho committed 212 days ago

b8862104

[pytree] support `X | Y` union type in `tree_map_only` (#120389)

XuehaiPan committed 212 days ago

be0ee934

[dtensor] implement scaled dot product attention (flash-attention) (#120298)

wanchaol committed 212 days ago

65627cfd

Revert "Native Half on ARM (#119483)"

pytorchmergebot committed 212 days ago

f2452e98

[ROCm] enable tests test_sampled_addmm_autograd_cuda_*, test_sample… (#117501)

dnikolaev-amd committed 212 days ago

c7328602

[DCP] Adds utility for converting torch save to dcp (#119815)

LucasLLC committed 212 days ago

1c1028ac

[FSDP2] disable compile in broken unit tests (#120358)

weifengpy committed 212 days ago

aae7ccd2

[DCP] Adds utility for converting dcp to torch save format (#119814)

LucasLLC committed 212 days ago

1ab441a7

[ROCm] Skip test_parity* unit tests in test_foreach only if ROCm version < 6.0 (#117301)

rraminen committed 212 days ago

e0a7b024

[inductor] Colorization improvements for bandwidth profiler (#120343)

bertmaher committed 212 days ago

de600508

Newer Older