strisunshinewentingwang/pytorch

Pull Requests Commits

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

Revert "Add ZT fastpath for remainder.Tensor JVP"

pytorchmergebot committed 2 years ago

da15d76e

[complex32] real and imag (also remove unused real and imag kernels)

kshitij12345 committed 2 years ago

0f1618ef

[PT-D][Sharding] Clean up sharded tensor code by leverage handle_torch_function (#76824)

fduwjj committed 2 years ago

4c504dd3

Add ZT fastpath for remainder.Tensor JVP

Nikita Karetnikov committed 2 years ago

7471f614

Allow creating new libraries and defining new operators from Python

anjali411 committed 2 years ago

07f766df

Allow users to override kernels for existing C++ ops through Python

anjali411 committed 2 years ago

55f55a4c

ci: Add upload credentials for rocm periodic test

seemethere committed 2 years ago

32ae5840

[ONNX] Format ONNX python with black

justinchuby committed 2 years ago

5dd1c677

Return all overloads for an operator in _jit_get_operation

ezyang committed 2 years ago

3a6da16a

Change TensorMeta to use tensor subclass.

ezyang committed 2 years ago

4a116783

Use TorchFunctionMode to implement PrimTorch tracing context

ezyang committed 2 years ago

48eb8d6a

[LT] Add a trie data structure for caching IR nodes

desertfire committed 2 years ago

f05710dd

Dispatch to mv rather than mm in the case that tensor1.ndim == 1 and tensor2.ndim == 2

lezcano committed 2 years ago

4baf7c08

Fix mv/addmv on CUDA when dealing with vectors of size=1 and stride=0

lezcano committed 2 years ago

02b5b92c

Refactor the API of the matmul implementation

lezcano committed 2 years ago

35ae9f68

[SR] Eliminate extra permutes around softmax calls (#76391)

Mike Iovine committed 2 years ago

1fed6b75

[SR] Codegen for aten::clamp (#76340)

Mike Iovine committed 2 years ago

cac2733a

Remove pow and float_power TestGradient Skips

drisspg committed 2 years ago

c59d5f17

Revert "Use scatter_reduce to support masked reductions on sparse COO tensors (sum, prod, amin, amax)"

pytorchmergebot committed 2 years ago

381e0830

[complex32] enable testing for multiple ops

kshitij12345 committed 2 years ago

8a2f207d

Newer Older