strisunshinewentingwang/pytorch

Pull Requests Commits

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

Sparse CSR: Add CPU fallback for sampled_addmm

IvanYashchuk committed 2 years ago

13355120

Added logit/reciprocal decomps, fixed var for complex, moved type promotion logic to standardize on primtorch's

Chillee committed 2 years ago

6917034a

Revert "Bugfix NAN and Inf handling for scatter_reduce (amin and amax)"

pytorchmergebot committed 2 years ago

4aa6b6b9

Fix unmarked fstring

peterbell10 committed 2 years ago

6df5a531

DOC fix momentum equation for nesterov

tomMoral committed 2 years ago

ff94c9de

Add high level control of fp32 matmul precision; disable TF32 for matmuls by default

eqy committed 2 years ago

e838137b

[ONNX] Support optional type (#68793) (#73284)

BowenBao committed 2 years ago

679fc90c

Fix false DeprecationWarning in `Module.state_dict`

lkct committed 2 years ago

b8776e14

[NNC] Lowering function generates the output buffer with the specified stride (#76529)

EikanWang committed 2 years ago

429a80dd

[ROCM] remove rocfft workaround

micmelesse committed 2 years ago

0878ba46

[Static Runtime] Add auto generated unstructured ops (#76398)

huiguoo committed 2 years ago

bcddd4ab

[Static Runtime] [RFC] Codegen support for ops with unstructured kernels (#76203)

huiguoo committed 2 years ago

ca0f2670

add timer for debugging long running tests

SherlockNoMad committed 2 years ago

734832e1

[SR] Fuse quantized linear/relu (#75775)

Mike Iovine committed 2 years ago

fc64dbdc

aten::_validate_sparse_compressed_tensor_args ops added to the ALLOW_LISST

b0noI committed 2 years ago

7dc13831

[FSDP] Optim state: ignore params if not in dict

awgu committed 2 years ago

94d65b05

Fixes prim type promotion and updates type promotion testing

Mike Ruberry committed 2 years ago

b557e102

Fix Rocm; Trigger build.

SherlockNoMad committed 2 years ago

65488723

Correct _validate_sparse_compressed_tensor_args signature

pearu committed 2 years ago

fc1f290e

switch Bazel to the shared generate-code genrule (#75790)

mikey dagitses committed 2 years ago

ac45fb9b

Newer Older