strisunshinewentingwang/pytorch

Pull Requests Commits

CoreML .mlmodel export support (#84784)

Bartek Rymkowski committed 2 years ago

0a6f3261

[NNC] add eltwise OPs: mish and elu (#80586)

chunyuan-w committed 2 years ago

ca419c33

Added additional simplifications/caching for replacements and divisibility (#84918)

Chillee committed 2 years ago

377b5d6f

[ONNX] Create decorators for symbolic function registration (#84709)

justinchuby committed 2 years ago

9d115523

Add symlink resolution in benchmark timer interface (#82734)

Mateusz Sypniewski committed 2 years ago

05ff3f89

Refs and decompositions for index_{add,copy,select,fill} (#85002)

lezcano committed 2 years ago

2f0b3de4

[ONNX] Update ONNX documentation to include unsupported operators (#84496)

justinchuby committed 2 years ago

d6c2080e

[ONNX] Update error messages (#85179)

justinchuby committed 2 years ago

46843be1

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

Newer Older