strisunshinewentingwang/pytorch

Pull Requests Commits

[dashboard] Remove aot_cudagraphs and nvprims_nvfuser

anijain2305 committed 1 year ago

cb891c90

Meta impl for linalg_cholesky and linalg_cholesky_ex (#89430)

anjali411 committed 1 year ago

9c0bf938

[quant][fx] Support producing reference quantized patterns for dynamic quantization (#89248)

jerryzh168 committed 1 year ago

c4e08387

[inductor] generate nan in the cpp backend (#89289)

desertfire committed 1 year ago

2823fc5e

[19/N] Add monitored_barrier custom op with CPU implementation (#89318)

H-Huang committed 1 year ago

5797f749

[18/N] Add allgather_coalesced custom op with CPU/CUDA implementations (#89317)

H-Huang committed 1 year ago

be22b5d3

Make aten.copy preserve strides (hf_Longformer) (#89464)

ezyang committed 1 year ago

d9cbe776

[Vulkan][TCC] Fix quantized shaders (#89456)

manuelcandales committed 1 year ago

2d94fd3b

Vectorized CPU code implementing right shift operator. (#88990)

alexsamardzic committed 1 year ago

0f7dca17

Reland Dispatch torch.norm to linalg.vector_norm and linalg.matrix_norm (#81761) (#84624)

lezcano committed 1 year ago

1d6a188d

[Checkpoint][2D][2/N] Add traverse for distributed checkpoint to core distributed (#89398)

wz337 committed 1 year ago

6b085d5c

Back out "[static-runtime] change the backend for permute_copy" (#89463)

Mike Iovine committed 1 year ago

7b0650d5

Revert submodule updates introduced by #89157 (#89449)

malfet committed 1 year ago

f2cf1b0f

Support masked_fill to address the GPT2 performance issue (#89274)

EikanWang committed 1 year ago

40cf214f

dynamo/torchxla integration: trace on xla rather than eager (#88904)

shunting314 committed 1 year ago

e545caa5

[Checkpoint][2D][1/N] Add dedup_tensors for distributed checkpoint to core distributed (#89399)

wz337 committed 1 year ago

1dae59ba

Fix retrying logic for successful unittest tests under --rerun-disabled-tests mode (#89454)

huydhn committed 1 year ago

ce342ed2

[vision hash update] update the pinned vision hash (#89471)

pytorchmergebot committed 1 year ago

338f6190

[PT-D][Tensor Parallelism][2/N] Sync TP API change to PT prod (#89467)

fduwjj committed 1 year ago

00b9473a

[inductor][compilation time] Fallback when kernel size for avg/max pool is large (#89448)

anijain2305 committed 1 year ago

82713a1c

[xnnpack][lite-int] Handle Constant Data (#89445)

mcr229 committed 1 year ago

496c8ae7

Revert "Added conv constraint that infers layouts (#89031)" (#89451)

anijain2305 committed 1 year ago

120d2006

dont clone symints, dont clobber symint proxies (#88230)

ezyang committed 1 year ago

06dffb33

[17/N] Add _reduce_scatter_base custom op with CPU/CUDA implementation (#88903)

H-Huang committed 1 year ago

58a74f34

Add torchvis support to dist bench (#89324)

wconstab committed 1 year ago

7174572b

Bind DispatchKey.Functionalonalize in pybind11 (#89452)

ezyang committed 1 year ago

57ed9480

[fix] tril & tril : out of bound check (#89384)

khushi-411 committed 1 year ago

b189a744

Mitigate flaky test_ops_fwd_gradients on macOS (#89410)

huydhn committed 1 year ago

dbc354b2

Suppress guards when creating fake tensors (#89349)

ezyang committed 1 year ago

ea50549c

Add commit hash to dynamo dashboard (#89462)

williamwen42 committed 1 year ago

fa4980cd

Older