strisunshinewentingwang/pytorch

Pull Requests Commits

[aot] fix disable amp for runtime wrapper

Valentine233 committed 2 years ago

ce7e1b2e

distinguish mutability of TensorImpl::data() (#97776)

mikey dagitses committed 2 years ago

2400cb1d

Removed hip call hipDeviceSynchronize (#97209)

BLOrange-AMD committed 2 years ago

6b9a1cf8

[primTorch] add ref for `cumprod` (#98670)

nkaretnikov committed 2 years ago

ff825de4

make TensorImpl::data_ptr_impl() non-const and have mutable in the name (#97744)

mikey dagitses committed 2 years ago

9d363616

Support LayerNorm without weight or bias parameters (#98687)

mrshenli committed 2 years ago

54b16848

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

mrshenli committed 2 years ago

1be3549a

Add overflow check for stride calculation (#94900)

CaoE committed 2 years ago

fdb04c6a

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

mingfeima committed 2 years ago

3925f6ed

ddp forward support custom backend. (#98283)

huihoaan committed 2 years ago

d95ee64b

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

nkaretnikov committed 2 years ago

a2e7910d

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

nkaretnikov committed 2 years ago

b411238d

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

mikey dagitses committed 2 years ago

387feaa1

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

shunting314 committed 2 years ago

2edfcafd

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

titaiwangms committed 2 years ago

d77d2f03

Restore CPU distributed tests (#97424)

huydhn committed 2 years ago

70535d60

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

huydhn committed 2 years ago

0fa25cbd

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

pytorchmergebot committed 2 years ago

cb3c4780

[ONNX] Refactor op level debugging (#97494)

titaiwangms committed 2 years ago

526d9bbc

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

Jiong Gong committed 2 years ago

5375e78b

[Inductor] cpp further code cleanup (#98135)

Jiong Gong committed 2 years ago

584a7ef3

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

eellison committed 2 years ago

85a90d91

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

yanboliang committed 2 years ago

a5f34686

reenable lowmem dropout (#98631)

Natalia Gimelshein committed 2 years ago

0dbdc8a3

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

mikey dagitses committed 2 years ago

c68a94c5

Add NLLLoss to DTensor prop rule (#98512)

mrshenli committed 2 years ago

d255c8e1

Set up automated hash pinning for triton (#97568)

clee2000 committed 2 years ago

a6155f34

Modify 'fake_tensor_unsupported' function (#98585)

jiayisunx committed 2 years ago

f959a0d5

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

peterbell10 committed 2 years ago

b7ff7172

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

peterbell10 committed 2 years ago

48397cdd

Older