strisunshinewentingwang/pytorch

Pull Requests Commits

clee2000 committed 1 year ago

f6cb4e89

clee2000 committed 1 year ago

08907f1c

Revert "Remove CUDA 11.6 from nightly builds (#93404)"

pytorchmergebot committed 1 year ago

f5e9c8ce

Revert "[inductor] fix crash issue when input is a view tensor (#90150)"

pytorchmergebot committed 1 year ago

5d259425

Basic Validation for FSDP `state_dict` transformations of modules with persistent buffers (#93396)

speediedan committed 1 year ago

769eca6f

Merge Inductor perf smoke test with other inductor CI tests (#93395)

desertfire committed 1 year ago

98e1b3e9

[inductor] Don't import torchvision (#93027)

peterbell10 committed 1 year ago

9ff7ddb2

[FSDP][3/N] Refactor `summon_full_params` unit tests (#92298)

awgu committed 1 year ago

481a334b

[FSDP][2/N] `_summon_full_params` -> `_unshard_params` (#92297)

awgu committed 1 year ago

10990734

Remove CUDA 11.6 from nightly builds (#93404)

atalman committed 1 year ago

c76ac8ee

Mark buffers that reuse other buffers (#93329)

wconstab committed 1 year ago

a14e3190

Refactor to allow reuse of SchedulerNode.allocate (#93328)

wconstab committed 1 year ago

d69876b2

retire sparse_mask_helper (#91714)

nikitaved committed 1 year ago

84187399

update fbgemm third party (#93907)

zhuhaozhe committed 1 year ago

a2fded30

[inductor] fix crash issue when input is a view tensor (#90150)

blzheng committed 1 year ago

b11ec270

[Inductor] add config for weight prepacking (#93811)

Valentine233 committed 1 year ago

a672fd1d

Check for none for NNModuleVariable.__module__ (#93326)

tugsbayasgalan committed 1 year ago

59ccc786

inductor: don't assert error when do cpu fx fusion for training mode (#93837)

XiaobingSuper committed 1 year ago

f4db47b1

inductor: separate bias from PackeLinear for better performance (#93348)

XiaobingSuper committed 1 year ago

3d020b69

[FSDP][optim_state_dict][10/N] Make optim_state_dict and optim_state_dict_to_load public (#92118)

fegin committed 1 year ago

4b0f1cc1

inductor: add conv+hardsigmoid fusion for cpu path(reland) (#93341)

XiaobingSuper committed 1 year ago

84ee50a2

[DTensor] implement dist_split as a sharding prop rule (#93306)

XilunWu committed 1 year ago

6f3018d5

[DTensor][fix] MultiThreadedTestCase misses _tls object and it won't reflect in CI (#93832)

XilunWu committed 1 year ago

966030f7

[DTensor] fix DTensorSpec dim_map description (#93160)

XilunWu committed 1 year ago

b82f93d5

inductor: align the decomposition output stride with none-decomposition path for torch.lerp (#93336)

XiaobingSuper committed 1 year ago

db873964

inductor: fix convert_shape_to_symint (#93349)

chunyuan-w committed 1 year ago

cff4d3bb

[PT-D][NamedOptimizer][6/N] Upstream init_state from keyed to NamedOptimizer (#93887)

fduwjj committed 1 year ago

e7ace1ff

[ROCm] Fix distributed tests failure and enable ROCm distributed CI (#92932)

jaglinux committed 1 year ago

f58ba553

Remove many untested dynamo backends (#93382)

jansel committed 1 year ago

569f2e32

[SDPA] Wire up FlashAttention's backward (#92917)

drisspg committed 1 year ago

653dc73d

Older