strisunshinewentingwang/pytorch

Pull Requests Commits

Cleanup Windows pip dependencies (#88862)

huydhn committed 1 year ago

6a757f1c

[MPS] Add floor_divide() op and its test case (#91126)

razarmehr committed 1 year ago

b63f0311

[FSDP][7/N] Support `replicate` in `fully_shard` (#91044)

awgu committed 1 year ago

aec09eeb

[FSDP][6/N] Add note explaining idioms for `_FSDPState` traversal (#90959)

awgu committed 1 year ago

e81ccfd1

[FSDP][5/N] Add manual "wrapping" support for `fully_shard` (#90874)

awgu committed 1 year ago

32fde537

[FSDP][4/N] Refactor func to share state/init handle attrs (#90871)

awgu committed 1 year ago

da9af986

Revert "use scatter_add for index_add when dim is the most inner dim (#88729)"

pytorchmergebot committed 1 year ago

3194281c

Remove debug code (#91148)

mlazos committed 1 year ago

07c340bb

Add cu118 workflows (#90826)

ptrblck committed 1 year ago

2d68cc4b

[dynamo] check buffers when checking accuracy (#91037)

williamwen42 committed 1 year ago

289f0643

Update patch release cherry pick condition (#90220)

albanD committed 1 year ago

17b80bfa

Bind autograd current Node for debugging purposes (#90867)

albanD committed 1 year ago

0eb45d54

use scatter_add for index_add when dim is the most inner dim (#88729)

CaoE committed 1 year ago

13dbad63

[CUDA12] Make PyTorch compatible with CUDA 12 (#91118)

jianyuh committed 1 year ago

63b8ecc4

Update dynamo xla test to make it part of the xla CI (#91130)

JackCaoG committed 1 year ago

7c58f1d4

[Checkpoint] Add test for fsdp model state saving and loading with/without resharding (#90950)

wz337 committed 1 year ago

29b119d0

fully_shard load state_dict (#90945)

rohan-varma committed 1 year ago

7330eabe

Revert "use libdevice for tanh (#90889)"

pytorchmergebot committed 1 year ago

95a115dd

[Dynamo] Fix builder for class with metaclass (#90807)

yanboliang committed 1 year ago

511fbad8

Revert "When nopython=True, Dynamo can't allow graph breaks. (#90970)"

pytorchmergebot committed 1 year ago

ef2bb9ca

Do not run inductor perf test with postnightly branch (#91133)

weiwangmeta committed 1 year ago

0f57e7f2

[Inductor] Replace graph.eliminate_dead_code() with graph.erase_node() in Permute Fusion (#91014)

jiawenliu64 committed 1 year ago

857ed2d7

[inductor] Rewrite Triton templates + epilogue fusion (retry) (#91105)

jansel committed 1 year ago

d6dd2e97

Removing invalid git option when updating submodules (#91132)

atalman committed 1 year ago

3bd37ff2

use libdevice for tanh (#90889)

ngimel committed 1 year ago

01488091

Fix non-existing parameters in docstrings in benchmarks (#91115)

kit1980 committed 1 year ago

30edd39b

Fix non-existing parameters in docstrings in misc places (#91121)

kit1980 committed 1 year ago

99bd8d12

Fix terminology within `linalg.slogdet` docs (#91129)

lezcano committed 1 year ago

0210d508

[Quant] lower fused LinearTanh for onednn backend (#89188)

Xia-Weiwen committed 1 year ago

a5eb564b

TorchDynamo: set output stride using eager output for cat (#89477)

XiaobingSuper committed 1 year ago

666d2180

Newer Older