strisunshinewentingwang/pytorch

Pull Requests Commits

[build] fix compilation error on s390x (#101923)

kiszk committed 1 year ago

9edf65a8

[MPS] Fix softplus with f16 input (#101948)

qqaatw committed 1 year ago

cce58a43

[dynamo][higher order op] Bugfixes to pass graph.lint (#102448)

anijain2305 committed 1 year ago

c3c14961

[ONNX] Don't duplicate model weights in ONNX export (#101134)

ilyasher committed 1 year ago

8d5b14e9

[benchmark] Flag to switch on activation checkpointing for HF models (#102557)

anijain2305 committed 1 year ago

33a49eea

Switch cuda 12.1 docker images to gcc9 (#102380)

atalman committed 1 year ago

6ac8a117

[Dynamo] Save global autocast state to restore on graph break (#102415)

yanboliang committed 1 year ago

9ff1932d

Revert "Quick fix for keep-going + reruns (#102569)"

pytorchmergebot committed 1 year ago

1a6ab8a5

[PT2][Quant][BE] refactor tets cose to reduce duplication and standardize (#102497)

kimishpatel committed 1 year ago

4f468646

Quick fix for keep-going + reruns (#102569)

clee2000 committed 1 year ago

7f6edcf4

[quant][pt2e] Add support for FixedQParamsQuantizationSpec (#102439)

jerryzh168 committed 1 year ago

f14ac74f

[fx] Fix repr when arg is an OpOverload (#102547)

anijain2305 committed 1 year ago

168ae806

[minifier] add missing import (#102521)

anijain2305 committed 1 year ago

68e55bff

Revert "[pt2] add `SymInt` support for `linalg.tensorsolve` (#102466)"

pytorchmergebot committed 1 year ago

95cdd58c

Revert "[pt2] add `SymInt` support for `linalg.vander` (#102469)"

pytorchmergebot committed 1 year ago

463df86c

Add type hints in torch/distributed/utils.py (#102262)

ringohoffman committed 1 year ago

c28f8e31

[pt2] add `SymInt` support for `linalg.vander` (#102469)

nkaretnikov committed 1 year ago

05717895

[pt2] add `SymInt` support for `linalg.tensorsolve` (#102466)

nkaretnikov committed 1 year ago

b1b76f61

[pt2] add `SymInt` support for `linalg.tensorinv` (#102465)

nkaretnikov committed 1 year ago

0ba81ce8

Add devcontainer support to PyTorch Project (#98252)

drisspg committed 1 year ago

7378b6b9

Move static checks of layers[0] (e.g., isinstance check) to model build time (#102045)

mikekgfb committed 1 year ago

4d89489d

DeviceMesh use dispatchable PG to support custom backend (#102336)

wanchaol committed 1 year ago

ff58d19c

[c10d] default backend need to check for nccl availability (#102470)

wanchaol committed 1 year ago

3ef4d697

implement __dir__ for dynamo (#102480)

alimoezzi committed 1 year ago

b02f48b1

Improve `clip_grad_norm` to use torch.linalg.vector_norm (#102429)

thomasw21 committed 1 year ago

704283d6

update triton pin (#101919)

Chillee committed 1 year ago

e71ab214

ASAN: fix heap-buffer-overflow (#101970)

AlekseiNikiforovIBM committed 1 year ago

5fa273c8

Fix silent nnz overflow for large sparse compressed tensors. (#102523)

pearu committed 1 year ago

fcbdbd66

Dynamo remaps legacy allgather to traceable one (#102232)

wconstab committed 1 year ago

77f97019

[inductor] Support multiple symbolic numel expr in CudaWrapperCodeGen (#102093)

desertfire committed 1 year ago

c58264c3

Older