strisunshinewentingwang/pytorch

Pull Requests Commits

Use deque instead of list for BFS (#91139)

kit1980 committed 1 year ago

83f4e30e

Add env var PYTORCH_TEST_RUN_EVERYTHING_IN_SERIAL=1 that allows running unit test suites in serial (#90981)

xwang233 committed 1 year ago

649d0b6a

Disable non-deterministic models for optimizers (#91149)

mlazos committed 1 year ago

2f5759ea

Update ProcessGroupRoundRobin (#91172)

H-Huang committed 1 year ago

f8b348c1

Don't run ios jobs on forks (#91112)

clee2000 committed 1 year ago

5ed5dfd9

nn/test_convolution to run in serial (#91113)

clee2000 committed 1 year ago

34717b3e

[cuDNN][cuDNN V8 API] (re-re-re-open) cuDNN V8 API on by default (#91117)

eqy committed 1 year ago

dabf515c

cleanup old python_compat code (#91162)

albanD committed 1 year ago

28ceccec

Set cmake PATH on macos to address libzstd flakiness (#91142)

huydhn committed 1 year ago

346fd040

[inductor] small CI improvements (#91140)

desertfire committed 1 year ago

84e73e12

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

Newer Older