strisunshinewentingwang/pytorch

Pull Requests Commits

Run test_torchinductor*.py with implicit_fallbacks=False (#94039)

Fabio Rocha committed 1 year ago

e116ca93

Pass input tensor __dict__ along to placeholder nodes (#94080)

voznesenskym committed 1 year ago

e44586a7

[BE] Modernize PyTorch even more for 3.8 with pyupgrade (#94520)

Skylion007 committed 1 year ago

9171f7d4

[SDPA] update type hint for scaled_dot_product_attention and documentation (#94008)

drisspg committed 1 year ago

70026aaa

Fix div by fp64 scalar issue on xla device (#94459)

ymwangg committed 1 year ago

9bef1ebb

Cleaning up some logic in tools/shared/cwrap_common.py (#94475)

joesho112358 committed 1 year ago

67513aee

add compile reason in InstructionTranslator RETURN_VALUE (#94176) (#94367)

zjjott committed 1 year ago

51cec7bf

[MPS] Fix cumsum for integral data types (#94530)

DenisVieriu97 committed 1 year ago

92d8c4b3

[fx] Fix matching args (#94375)

angelayi committed 1 year ago

d990ddad

fix: forbid multi-index for index_select over scalar (#94347)

ganler committed 1 year ago

db6cfff8

feature: adding the ability to restore shapes after loading a traced model (#90744)

mnuyens committed 1 year ago

0d0ebcdf

Fix bug in unsqueeze_nested stride calculation (#88688)

mikaylagawarecki committed 1 year ago

c7c72389

[ONNX] Skip import test for experimental files (#94552)

BowenBao committed 1 year ago

889a4640

port sparse_mm.reduce to pytorch and optimize it on CPU (#83727)

mingfeima committed 1 year ago

c620ece7

Add config option to reduce warnings in inductor (#94413)

jansel committed 1 year ago

24ae50bc

[MPS] Fix min/max_reduction_with_dim ops (#94386)

kulinseth committed 1 year ago

1d398065

[MPS] Add im2col and col2im to Fallback (#94491)

kulinseth committed 1 year ago

0fe11589

WelfordOps: Remove combine_t and use acc_scalar_t instead (#94522)

peterbell10 committed 1 year ago

a21bddcc

[decomp] Use var_mean in native_batch_norm decomposition (#94140)

peterbell10 committed 1 year ago

e22e323b

Fix embedding_dense_backward to not cast indiices to floats (#94572)

Chillee committed 1 year ago

e844120b

Don't throw tf32 warning if no nodes in graph are matmuls + fp32 + cuda (#94561)

Chillee committed 1 year ago

1770ccf6

Revert "update aten op overload to not use `from` to avoid compile errors (#89797)"

pytorchmergebot committed 1 year ago

f152a79b

teach inductor to handle floor (#94341)

ngimel committed 1 year ago

a5daea69

inductor: don't do transpose vectoriztion if input ld depends on most inner var (#94493)

XiaobingSuper committed 1 year ago

02b8a7f4

Renamed passes to options in torch.compile (#94500)

Chillee committed 1 year ago

3a12b16f

[MPS] Fix the Channels last bug with GradientWithInput. (#94384)

kulinseth committed 1 year ago

59e87566

[MPS] Casting int64 to int32 for reduction ops and raise warning. (#94484)

kulinseth committed 1 year ago

8dbe63c9

don't call floor for symint unless necessary (#94365)

ngimel committed 1 year ago

715f3733

Enable Python-3.11 binary builds across the board (#94430)

malfet committed 1 year ago

89df0e42

[MPS] Fix batchnorm forward and backward pass (#94351)

DenisVieriu97 committed 1 year ago

a1f15fb9

Newer Older