strisunshinewentingwang/pytorch

Pull Requests Commits

angelayi committed 2 years ago

64fb268b

angelayi committed 2 years ago

aaf7b052

angelayi committed 2 years ago

e7a39780

angelayi committed 2 years ago

929e9799

address comments

angelayi committed 2 years ago

4797cf19

[dynamo] Fix Autograd Function Classmethod bug

angelayi committed 2 years ago

a0039e95

[export] Initial deserialization v2 (#102716)

angelayi committed 2 years ago

e930c0fc

insert to dtype for fused mem copy scheduler node (#101042)

zhuhaozhe committed 2 years ago

adcefcb3

Fix graph break on boolean mask better (#103052)

ezyang committed 2 years ago

605a8524

Add a little more error checking to minifier (#103057)

ezyang committed 2 years ago

2dafa70d

add foreach support for custom device (#102047)

heidongxianhua committed 2 years ago

e4a42bcf

[c10d] Make it default that PG do not perform barrier after init (#103033)

kwen2501 committed 2 years ago

07104ca9

update argument checks from padding layers (#102253)

mingfeima committed 2 years ago

3e988316

[vision hash update] update the pinned vision hash (#103120)

pytorchupdatebot committed 2 years ago

5acf7e26

[FSDP][1/N]Add device_mesh to FSDPstate (#102317) (#102551)

wz337 committed 2 years ago

a02a58d8

Disable dynamo on some opt methods and differentiable optimizer tests (#103066)

mlazos committed 2 years ago

0769a50a

Teach Triton codegen to generate sqrt (#103084)

ezyang committed 2 years ago

f7608998

[PT-D] Update torch.distributed code owners (#103114)

fduwjj committed 2 years ago

3f6f5086

Back out "Remove `check` from `_prims_common`, replace with `torch._check*` (#102219)", Back out "Forwatd fix for D46427687" (#103128)

izaitsevfb committed 2 years ago

82149371

[benchmarks] Torchbench llama is not suitable for training (#103094)

anijain2305 committed 2 years ago

428bff84

Add device range helper and remove sm86 specific check for memory efficient attention (#102985)

drisspg committed 2 years ago

2800a04a

[export] Remove example custom_object_type to type_reflection_method. (#103015)

zhxchen17 committed 2 years ago

6596cfa4

[ONNX] Add FX exporter MaxPool tests (#102773)

titaiwangms committed 2 years ago

27f4dc6c

Disable fallback for custom kernels (#101131)

zou3519 committed 2 years ago

5b700fc9

[PT2][Quant] Move embedding quantization to osss (#103088)

kimishpatel committed 2 years ago

8e0837cf

[inductor] add a few tests to verify view_to_reshape pass is safe (#103034)

shunting314 committed 2 years ago

bf312f2d

[Dynamo] No graph break for super(MyConv{1/2/3}d, self).forward and super(MyConvTranspose, self).forward (#102509)

weifengpy committed 2 years ago

61736679

torch.compile docs: "Profiling to understand torch.compile performance (#102862)

davidberard98 committed 2 years ago

038955f4

dst_bin_of_end_center is defined twice (#102755)

lonelyeagle committed 2 years ago

62610554

[Profiler] Update Kineto Submodule (#103031)

aaronenyeshi committed 2 years ago

0279d0b6

Older