strisunshinewentingwang/pytorch

Pull Requests Commits

[dynamo] Support module backwards hooks (#120685)

jansel committed 204 days ago

e3dbd194

[dynamo] Fix convolution meta kernel when input channel is 0 (#120944)

xmfan committed 204 days ago

9b2c35b4

Reinplace auto_functionalized (#120829)

zou3519 committed 204 days ago

d534a497

[Composable APIs] Add composable API `fully_shard` deprecation warning (#120929)

wz337 committed 204 days ago

791f8ef3

Teach dynamo about vjp (#119405)

guilhermeleobas committed 204 days ago

fd35aafc

[DCP] Adds storage reader and planner classes for online loading/sharding of models in torch.save format (#119816)

LucasLLC committed 204 days ago

9d5dea78

Revert "Fix guard for SUPPORTED_NODES (#120798)"

pytorchmergebot committed 204 days ago

33da8d5c

Fix nll loss dynamo failure (#120805)

eellison committed 204 days ago

7ebfe217

Pass inductor strides forward in ddp optimizer (#120523)

eellison committed 204 days ago

d03b11ad

Fix handling of torch.return_types in dynamo (#120826)

mreso committed 204 days ago

772db2a3

Fix isin decomp and add python meta registration (#120821)

janeyx99 committed 204 days ago

da559c98

Revert "[C10D] Add ProcessGroup op_id to track ops inside coalescing region (#120745)"

pytorchmergebot committed 204 days ago

76d3a6bb

[dynamo][easy] Dynamo test changes (#120927)

anijain2305 committed 204 days ago

e7039e3a

Skip semi-structured-sparse on windows (#120807)

drisspg committed 204 days ago

39c092d2

[rocm][cmake] retrieve rocm location from ROCM_SOURCE_DIR env if specified (#120898)

chenyang78 committed 204 days ago

1a1f58ff

[FSDP2][DCP][DSD] Add test to ensure FSDP2 model/optim state_dict work after a full training loop (#120871)

wz337 committed 204 days ago

b2dddcfe

[TorchElastic] Refactoring to support non-default logging strategy (#120691)

kurman committed 204 days ago

67d3e4f2

[FSDP2][ez] Combined communication test files (#120904)

awgu committed 204 days ago

277bc977

Revert "[inductor] Optimize welford reduction (#120330)"

pytorchmergebot committed 204 days ago

0b924d7c

SymIntify prod_backward (#120776)

ezyang committed 204 days ago

0a766680

[c10d] fix the unwanted reason (#120863)

shuqiangzhang committed 204 days ago

313abcdb

Refine value ranges on inequalities (#120800)

ezyang committed 204 days ago

f94933ed

[functional collecitve] don't import torchdynamo when running torchdeploy (#120900)

yifuwang committed 204 days ago

81c4c0dd

fix dupe deprecated warning in dynamo export (#120896)

avikchaudhuri committed 204 days ago

f7a809c9

Test TD (test removal) on crossref (#119426)

clee2000 committed 204 days ago

0290fe65

Revert "Update flash_attention kernel from 2.3.6 to 2.5.5 (#118935)"

pytorchmergebot committed 204 days ago

1458f1de

[inductor max autotune] Detailed autotuning result logs ( machine-readable ) (#119004)

kadeng committed 204 days ago

96eff4ef

[dynamo] Graph break when faking named tensors (#120779)

jamesjwu committed 204 days ago

a911eb74

[Dynamo] Fix inspect.getattr_static doesn't work well for torch.utils._cxx_pytree.PyTreeSpec (#120812)

yanboliang committed 204 days ago

1104e079

[rocm][cmake] correctly check the ROCM_SOURCE_DIR environment (#120858)

chenyang78 committed 204 days ago

ca679384

Newer Older