strisunshinewentingwang/pytorch

Pull Requests Commits

[cuDNN] Change `cuDNN` context manager to enable cuDNN as default

eqy committed 1 year ago

Verified 2fd6c33b

Fix FastToLocals call by copy pasting (#91168)

albanD committed 1 year ago

c7f1974c

Fix all simple compilation issues in eval_frame.c (#91166)

albanD committed 1 year ago

5e77971a

Upgrade lintrunner numpy to a version supported by 3.11 (#91164)

albanD committed 1 year ago

b7f48d71

Use python compat from python/pythoncapi_compat (#91163)

albanD committed 1 year ago

c0e7d8f8

Revert "[MPS] Add `group_norm[fwd+backward]` and `mean_var` (#91190)"

pytorchmergebot committed 1 year ago

645eda0a

[cuBLAS] Add an option to disable reduced precision reductions for BF16 GEMM (#89172)

eqy committed 1 year ago

8b617f81

Temporarily disable ROCm periodic tests (#91256)

jeffdaily committed 1 year ago

1c7e8157

[MPS] Add MPSHooks interface to enable accessing MPS functions globally (#91104)

razarmehr committed 1 year ago

eeb9154b

[MPS] Add `group_norm[fwd+backward]` and `mean_var` (#91190)

malfet committed 1 year ago

371716eb

Don't graph break on patched module methods or aliased methods (#91018)

williamwen42 committed 1 year ago

d6fc2d82

Dynamo, FX, Inductor Progress Bars (#88384)

msaroufim committed 1 year ago

15af4b1c

Compile fix for Clang + libc++ (#91212)

splhack committed 1 year ago

bfdc0358

[Re-landing 86706] [JIT] Frozen Graph Linear-BatchNormNd Folding (#91020)

min-jean-cho committed 1 year ago

6d2b0cbb

Integrate apply_optim_in_backward with DDP (#89194)

rohan-varma committed 1 year ago

e8bf7c21

[inductor] Update how REQUIRE_HIGHER_TOLERANCE is handled (#91227)

desertfire committed 1 year ago

8992eec7

[MPS] Fix index_add with non-f32 inputs (#88542)

qqaatw committed 1 year ago

b7f35e41

Update debug option for torch._dynamo (#91223)

take-cheeze committed 1 year ago

04762014

[autograd Function] Don't materialize forward grad for non-differentiable types (#91183)

soulitzer committed 1 year ago

b66862ba

Store source, not sname, in Symbol (#91057)

ezyang committed 1 year ago

88c581be

Update torchrun and TorchElastic to take optional `local_addr` param to allow skip local IP lookup if specified (#88922)

cz-37 committed 1 year ago

5d37890b

Restructure ShapeEnv so it uses GuardBuilder.SHAPE_ENV directly (#91055)

ezyang committed 1 year ago

57390116

[torchgen] Add CI job to make sure torchgen works for Executorch op registration (#89596)

larryliu0820 committed 1 year ago

2f154f68

[QNNPACK] Add more unaligned attributes (#91208)

digantdesai committed 1 year ago

37ea99cd

[MPS] Fix data type issues in Unary ops (#91120)

razarmehr committed 1 year ago

a274b5b9

[BE] Use `aten` global in `torch._refs` (#91189)

malfet committed 1 year ago

c8546c93

[BE] Use `aten` global var (#91188)

malfet committed 1 year ago

46f64117

[MPS] Fix `torch.std`/`torch.var` default/correction handling (#91203)

malfet committed 1 year ago

dd735b96

Decompose fill, zero, and zeros_like (#90968)

peterbell10 committed 1 year ago

e670c261

Temporarily disable ROCm tests (#91217)

kit1980 committed 1 year ago

eeacb6ae

Older