strisunshinewentingwang/pytorch

Pull Requests Commits

Enable replicated embedding in SPMD for NLP models (#98686)

mrshenli committed 1 year ago

1be3549a

Add overflow check for stride calculation (#94900)

CaoE committed 1 year ago

fdb04c6a

add Half to cat fast path on CPU (#96078)

mingfeima committed 1 year ago

3925f6ed

ddp forward support custom backend. (#98283)

huihoaan committed 1 year ago

d95ee64b

[pt2] remove skip for `masked.logsumexp` in `test_proxy_tensor.py` (#98676)

nkaretnikov committed 1 year ago

a2e7910d

[pt2] add meta function for `logcumsumexp` (#98683)

nkaretnikov committed 1 year ago

b411238d

add mutable to name of non-const Storage::data_ptr (#97694)

mikey dagitses committed 1 year ago

387feaa1

[inductor] remove RBLOCK from persistent reduction kernel's parameter list (#98653)

shunting314 committed 1 year ago

2edfcafd

[ONNX] Fix scalar elements in op.Concat (#98509)

titaiwangms committed 1 year ago

d77d2f03

Restore CPU distributed tests (#97424)

huydhn committed 1 year ago

70535d60

Fix broken MacOS build due to #97690 (#98665)

huydhn committed 1 year ago

0fa25cbd

Revert "refactor(add privateuseone floder in aten/src/ATen): add a PrivateUse… (#98127)"

pytorchmergebot committed 1 year ago

cb3c4780

[ONNX] Refactor op level debugging (#97494)

titaiwangms committed 1 year ago

526d9bbc

[Inductor] turn on vectorization with fallback for indirect indexing etc. (#98138)

jgong5 committed 1 year ago

5375e78b

[Inductor] cpp further code cleanup (#98135)

jgong5 committed 1 year ago

584a7ef3

Rename assert options, turn off by default (#98616)

eellison committed 1 year ago

85a90d91

[Dynamo] Fix bug when dynamo generate guards for enum type (#98652)

yanboliang committed 1 year ago

a5f34686

reenable lowmem dropout (#98631)

ngimel committed 1 year ago

0dbdc8a3

distinguish mutability of untyped Storage::data (#97690)

mikey dagitses committed 1 year ago

c68a94c5

Add NLLLoss to DTensor prop rule (#98512)

mrshenli committed 1 year ago

d255c8e1

Set up automated hash pinning for triton (#97568)

clee2000 committed 1 year ago

a6155f34

Modify 'fake_tensor_unsupported' function (#98585)

jiayisunx committed 1 year ago

f959a0d5

[inductor] Use 64-bit indexing for large tensors in triton code (#97447)

peterbell10 committed 1 year ago

b7ff7172

[inductor] Fix benchmark_compiled_module codegen with CppWrapperCodeGen (#98608)

peterbell10 committed 1 year ago

48397cdd

Fix pytest config (#98607)

peterbell10 committed 1 year ago

917e9f11

Update the use of nvidia-smi for GPU healthcheck (#98036)

huydhn committed 1 year ago

4563adac

Back out "[kineto] add SOFT_ASSERT when logging metdata" (#98630)

christ1ne committed 1 year ago

112dfa14

[Dynamo] Include UserDict in clone_inputs (#97725)

YJ Shi committed 1 year ago

5ceae85f

Add comments about maybe_guard usage in Inductor (#98563)

ezyang committed 1 year ago

cf10fd82

Back out "`GradScaler` recomputes `optimizer_state["found_inf_per_device"]` before `optimizer.step` (#97415)" (#98613)

apeeyush committed 1 year ago

ebd4c165

Newer Older