strisunshinewentingwang/pytorch

Pull Requests Commits

[FSDP] Relax `sharded_grad` assert to allow IDLE (#96584)

awgu committed 1 year ago

a916d649

bsr_dense_mm Triton kernel: fix out kwarg (#96648)

nikitaved committed 1 year ago

05dda7ff

[AO] Update qLSTM implementation to remove unsupported backend ops (#96436)

Ninja91 committed 1 year ago

40df3b41

Moves SDPA backward helper native function to functionsmanual.cpp (#95821)

drisspg committed 1 year ago

7ec0d6f0

Add tests for all padding layers to `module_db` in `common_modules.py` (#96641)

rishub-tamirisa committed 1 year ago

152c1529

Disable flaky linalg.det.singular tests on ROCm (#96707)

huydhn committed 1 year ago

4562898a

Add missing ceil for libdevice in triton (#96709)

ezyang committed 1 year ago

0d3bf2fd

[fix] resize_, resize_as_ : version bump in ADInplaceOrView (#96598)

kshitij12345 committed 1 year ago

1ec65556

[reland2][inductor] Add an AOT compilation mode for Inductor CPP backend (#96520)

desertfire committed 1 year ago

f03db8d6

debug shape guards (#95848)

avikchaudhuri committed 1 year ago

178d2a38

[Inductor] Enable Inductor to support BF16 atomic_add (#96620)

EikanWang committed 1 year ago

a37197df

Correctly use PythonPrinter for generating wrapper code referencing sympy (#96710)

ezyang committed 1 year ago

ff7e510d

update_expected.py to parse artifacts and update graph break stats (#96480)

wconstab committed 1 year ago

f1d4d291

Suport tensor type for XPU (#96656)

guangyey committed 1 year ago

9239279c

Revert "[MPS] Add C++ API support for MPS backend (#96668)"

pytorchmergebot committed 1 year ago

a07817ad

[Inductor] Legalize BF16 (#96183)

EikanWang committed 1 year ago

bdd09e68

[Inductor] apply vec float mask on logical comparison ops in cpp (#96502)

jgong5 committed 1 year ago

190e284b

[Inductor] Fix the incorrect fusion if a Conv/Linear moduel is called from multiple places (#96485)

EikanWang committed 1 year ago

3f723546

[Inductor] Skip the hf_T5_base due to intermittent failure on CI (#96649)

EikanWang committed 1 year ago

3cad8d23

control_flow.{cond/map} allows tracked_fakes divergence (#96546)

yanboliang committed 1 year ago

166117e0

[primTorch] add meta implementation for `upsample_nearest2d_backward` (#96612)

nkaretnikov committed 1 year ago

ec536232

Move all ONNX test dependencies to Docker (#96590)

huydhn committed 1 year ago

6a2dcfd7

[CUDA] Abate spurious resize warnings in MultiMarginLoss backward (#96382)

eqy committed 1 year ago

70090b4d

[DDP] Enable delayed all reduce in DDP (#96673)

GD06 committed 1 year ago

906a1952

[vision hash update] update the pinned vision hash (#96703)

pytorchmergebot committed 1 year ago

d0a4881d

Add extra metadata (as comments) to Inductor generated code (#96581)

Chillee committed 1 year ago

2a08a627

Fix calls to sizes to enable dynamic shapes with sdpa (#96674)

drisspg committed 1 year ago

f56cb41c

Check dynamo graph-breaks in CI (#96346)

wconstab committed 1 year ago

218eeaca

Clean up duplicated retry function in common.sh (#96682)

huydhn committed 1 year ago

2cc8368a

[BE] Remove fast_nvcc tool (#96665)

malfet committed 1 year ago

62c1e33f

Newer Older