strisunshinewentingwang/pytorch

Pull Requests Commits

[PyTorch] Make RecordFunction store inputs as ArrayRef (#72484)

swolchok committed 2 years ago

52af4fc5

Introduce distributed checkpoint with ShardedTensor.

Rodrigo Kumpera committed 2 years ago

710246ea

Bugfix scatter_reduce backward formulas

mikaylagawarecki committed 2 years ago

465e0ae2

Load state dict post hook

rohan-varma committed 2 years ago

56bed0dc

Deprecate torch.lu_solve

lezcano committed 2 years ago

f84d4d9c

Deprecate torch.lu

lezcano committed 2 years ago

a5bbfd94

Update torch.lu_unpack docs

lezcano committed 2 years ago

9dc8f256

Add linalg.lu_solve

lezcano committed 2 years ago

fc5b4a5a

[RPC small change] Improving logging for store.wait error

H-Huang committed 2 years ago

31d03c2f

Fix rpc_test.py MyConvNetForMNIST location and sparse tests

pbelevich committed 2 years ago

0a770f80

Handle torch.memory_format serialization in TensorProperties.

Rodrigo Kumpera committed 2 years ago

55e90d17

Fix issue with _checkpoint_without_reentrant

jamaliki committed 2 years ago

bc3c7a6c

[SR] Use at::DimVector in reshape_copy_out (#76473)

Mike Iovine committed 2 years ago

9e32cded

[ROCm][GHA] split 4 GPU hosts across two runners

jeffdaily committed 2 years ago

61aec161

[Reland take-2] Add JIT graph fuser for oneDNN Graph API (v0.5)

sanchitintel committed 2 years ago

4ee29d60

exclude distributed tests from windows

clee2000 committed 2 years ago

94fc92f2

Adds amax and amin references

ngimel committed 2 years ago

1c776d20

Port clamp_min and clamp_max to structured

ngimel committed 2 years ago

12279891

rohan-varma committed 2 years ago

b30c027a

Minor follow up fixes for python registration

anjali411 committed 2 years ago

3d28ab07

ns for fx: skip shadowing ops if copy subgraph is not implemented (#76663)

vkuzo committed 2 years ago

3a8752db

ns for fx: skip shadowing for torch.cat, and also for nodes with only kwargs (#76561)

vkuzo committed 2 years ago

d3e33893

[FSDP] Include buffers in `ignored_modules`

awgu committed 2 years ago

73b33de9

`functional.max_unpool`: OpInfo tests + simpler backward + forward ad + fwad over backward ad

nikitaved committed 2 years ago

33fabe9a

Go through the dispatcher in matmul_out for the 1D-1D case

lezcano committed 2 years ago

2b3b205f

lezcano committed 2 years ago

7cb7cd58

Improve derivative of QR decomposition

lezcano committed 2 years ago

1a4eea57

Sparse CSR: Fix sampled_addmm for noncontiguous inputs and fix block sparse triangular solve

IvanYashchuk committed 2 years ago

3df0140c

anjali411 committed 2 years ago

ad8386c1

[lint] preoperly suggest lintrunner when lint fails

suo committed 2 years ago

6a44e066

Newer Older