strisunshinewentingwang/pytorch

Pull Requests Commits

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

[MPS] Calculate nonzero count inside nonzero op (#94442)

DenisVieriu97 committed 1 year ago

016f0b2f

[Profiler] Use RAII wrapper to manage refcounts during python tracer startup. (#91646)

Taylor Robie committed 1 year ago

4c6a7fae

[MPS] Enable index add for TestConsistency (#94356)

DenisVieriu97 committed 1 year ago

336d9354

[MPS] Add the floor_divide fixes. (#94488)

kulinseth committed 1 year ago

299ada9c

Fix saved tensor hooks to propogate errors back to python as-is (#94456)

soulitzer committed 1 year ago

93d7d546

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

desertfire committed 1 year ago

2a585173

Mark ROCm trunk job as unstable (#94550)

huydhn committed 1 year ago

79ed6b24

[quant][fx] Change prepare_fx and convert_fx to preserve the GraphModule type of input (#94412)

jerryzh168 committed 1 year ago

2394e6ba

[dtensor] update readme for prototype release (#94517)

wanchaol committed 1 year ago

09598b60

[ROCm] Remove PYTORCH_MIOPEN_SUGGEST_NHWC flag (#90725)

jeffdaily committed 1 year ago

66bfcd32

ao migration: fix broken import, try 2 (#94458)

vkuzo committed 1 year ago

c1e27046

Newer Older