strisunshinewentingwang/pytorch

Pull Requests Commits

[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

[MPS] Fix addmm calculation (#94534)

DenisVieriu97 committed 1 year ago

2ad29009

Revert "Set torch.backends.cudnn.enabled to false when testing accuracy (#94363)"

pytorchmergebot committed 1 year ago

10c430ba

Add exception handlers for stoll in jit/frontend/schema_type_parser.cpp (#94295)

m4drat committed 1 year ago

a1d210de

Assert TensorBox produced by lowering and add [Note: Inductor IR] (#94361)

wconstab committed 1 year ago

d21a7e71

add mixed data type support for LayerNorm backward on CPU (#88064)

jiayisunx committed 1 year ago

01de5dda

Dynamo Export use fake tensor (#94276)

SherlockNoMad committed 1 year ago

54fa9801

Lower libtorch build parallelization to avoid OOM (#94548)

huydhn committed 1 year ago

2af89e96

Add uint8 support for interpolate for CPU images (#90771)

NicolasHug committed 1 year ago

544c04f2

[quant] Add quantize and dequantize operators to decomposition table (#93312)

jerryzh168 committed 1 year ago

782e4f5c

small bugfixes to release notes script (#94536)

mikaylagawarecki committed 1 year ago

df13247e

[inductor] Fix a conv stride assertion (#94405)

desertfire committed 1 year ago

93ee1bf1

Ignore 7z locked usage log error on Windows non-ephemeral runners (#94483)

huydhn committed 1 year ago

f5ccbc17

Newer Older