strisunshinewentingwang/pytorch

Pull Requests Commits

Add a missing argument (#108477)

cyyever committed 1 year ago

aff7fdcb

[aten decomp] Update sdpa decom (#108371)

kimishpatel committed 1 year ago

cc50e654

[Doc] Update the dynamo deepdive doc (#108147)

youkaichao committed 1 year ago

ba9acbeb

Use std::filesystem in c10 tempfile and tempdir (#106656)

cyyever committed 1 year ago

7b91f762

Use contiguous() to handle noncontiguous outputs during elementwise decomposition (#108140)

VRSinghHabana committed 1 year ago

1b3dc05c

NT support for cat with dim > 0 when representable as jagged (#108428)

jbschlosser committed 1 year ago

e5548f81

NT support for narrow() on dim=0 (#108362)

jbschlosser committed 1 year ago

76ccf6c7

[gen_operators_yaml] add arguments to control include_all_overloads (#108396)

stephenyan1231 committed 1 year ago

01b662ba

[AOTInductor][Reland] Proxy Executor for Extern Fallback kernels (#107279) (#108350)

SherlockNoMad committed 1 year ago

b9dfdc09

[Dynamo] Update the implementation of _debug_get_cache_entry_list (#108335)

youkaichao committed 1 year ago

b9fc6d7d

Improve docs for `torch.unique` `dim` argument (#108292)

kurtamohler committed 1 year ago

de586001

[Reland] Improve MKL related logic in FindOpenMP.cmake (#104224)

cyyever committed 1 year ago

0cc2f06a

[Quantization] Add metadata porting for nodes added by quantization (#107107)

kimishpatel committed 1 year ago

ffc0c460

[BC BREAKING] Remove outdated python submodules (#108236)

cyyever committed 1 year ago

d6a9c2b4

[Quant] Add DQ duplication pass (#107900)

kimishpatel committed 1 year ago

eb67c452

[Quant] Bug fix (#107899)

kimishpatel committed 1 year ago

f8d1ca98

[Quantization] Make annotation util functions return annotated nodes (#107106)

kimishpatel committed 1 year ago

37b0d76e

[Quant] Use input_qspec_map for weight quantization of linear (#107105)

kimishpatel committed 1 year ago

99168c1f

[vision hash update] update the pinned vision hash (#108460)

pytorchupdatebot committed 1 year ago

ab6a86dc

Refactorings for constant folding (#108450)

eellison committed 1 year ago

ed92d934

do not cast all inputs in benchmarks (#108456)

msaroufim committed 1 year ago

5f5caed2

[CUDACaching Allocator] Release the allocator lock on the slow path (#108367)

banitag1 committed 1 year ago

b8af8ac7

Only add triton dependency to CUDA and ROCm binaries if it hasn't been set as an installation requirement yet (#108424)

huydhn committed 1 year ago

4084d039

Add dynamo support for `rdiv` dunder method. (#108422)

ysiraichi committed 1 year ago

2e3fce54

[inductor] Handle aten.full's dtype in the decomposition (#108443)

peterbell10 committed 1 year ago

fa8edd93

Revert "Horizontally fuse input concatenation (#108115)"

pytorchmergebot committed 1 year ago

2c1f0772

[S362716] turn off constant folding (#108389)

xw285cornell committed 1 year ago

a27f0108

Make make_fx cond preserve node meta (#108356)

ydwu4 committed 1 year ago

e3933609

[pt2] Turn on cudagraph tree in fbcode (#108416)

xw285cornell committed 1 year ago

ac42b4ea

print equalities (#108427)

avikchaudhuri committed 1 year ago

ad032a76

Newer Older