strisunshinewentingwang/pytorch

Pull Requests Commits

Retry installing lintrunner if download fails (#85189)

ZainRizvi committed 2 years ago

a4c7cadc

Revert "[FSDP] Option to keep grads in lower prec (#85134)"

pytorchmergebot committed 2 years ago

14b3bdc0

Remove assertEqualIgnoreType from test_pooling (#85112)

mshuaibii committed 2 years ago

4382da5d

[ONNX] New symbolic function registry (#84382)

justinchuby committed 2 years ago

cd7e6d4a

update torch.narrow doc (#85180)

kshitij12345 committed 2 years ago

73515435

fix for rebase and merge (#85168)

clee2000 committed 2 years ago

5877cc9a

[functorch] Document DynamicLayer.{h, cpp} a bit more (#85178)

zou3519 committed 2 years ago

17593f15

[QNNPACK] Export cpuinfo-targets in clog CMakeLists (#84876)

salilsdesai committed 2 years ago

d559299c

[FSDP] Short-term fix to remove `optim_input` (#84201)

awgu committed 2 years ago

c6c3346d

[Testing] Port `bernoulli` and `multinomial` to ErrorInputs. (#74683)

khushi-411 committed 2 years ago

a9258eba

[functorch] remove argnums partial helper function, rewrite test to use slice argnum (#84951)

samdow committed 2 years ago

a5d9d2aa

Revert "Make ones and zeros's ref accepts variadic size argument (#85117)"

pytorchmergebot committed 2 years ago

776e0fe7

New calling convention for Python dispatcher (#85133)

ezyang committed 2 years ago

490727a3

Optimize torch.ops.ns.opname.overload accessor in torch dispatch (#85132)

ezyang committed 2 years ago

e5fac7f5

[FSDP] Option to keep grads in lower prec (#85134)

rohan-varma committed 2 years ago

607eccb1

Make ones and zeros's ref accepts variadic size argument (#85117)

SherlockNoMad committed 2 years ago

7e5616c9

flash_attention_helper mitigation: pass contiguous inputs (#85135)

cpuhrsch committed 2 years ago

38778add

Increase docker build timeout (#85156)

ZainRizvi committed 2 years ago

7b3e177b

Use alias for nop decomp (#84727)

SherlockNoMad committed 2 years ago

29eba319

Rename 'torch/ao/nn/quantized._reference' to 'torch/ao/nn/quantized/reference'. (#84974)

fufeisi committed 2 years ago

d8eae628

Implement forward AD for scatter_reduce (#85000)

lezcano committed 2 years ago

d710c95c

fix half_to_float arg in *softmax decomp (#85120)

ngimel committed 2 years ago

6162a043

Re-enable fixed dynamo tests (#84969)

eellison committed 2 years ago

f37069aa

Flip fake tensors on in aot autograd (#84968)

eellison committed 2 years ago

54c9c4e7

Add warning about installing functorch via setup.py (#85095)

zou3519 committed 2 years ago

61ba1250

Re-enables some tests for linalg.det (#85141)

lezcano committed 2 years ago

2e1ec5d1

Fix behaviour of index_add / atomicAdd(bool,bool) (#85100)

lezcano committed 2 years ago

8b29b795

Added support for symbolic is_contiguous (#84829)

Chillee committed 2 years ago

4bdc0af5

[FSDP] Add `_set_flattened()`; `_is_flattened()` (#85038)

awgu committed 2 years ago

5652ab22

[vision hash update] update the pinned vision hash (#85130)

pytorchmergebot committed 2 years ago

0ec19db7

Newer Older