strisunshinewentingwang/pytorch

Pull Requests Commits

inductor: add fake mode tracing for cumsum graph pattern (#102820)

XiaobingSuper committed 1 year ago

ca180539

enforece dtype (#102802)

uciyc123 committed 1 year ago

8e2a86c2

[inductor] Fix a cpp_wrapper issue when fx_passes modified fx graph (#102851)

desertfire committed 1 year ago

881307ab

[export] Replicate exportdb examples and tests in oss. (#102769)

zhxchen17 committed 1 year ago

26bf8894

[CheckpointWrapper] Warn on reentrant use (#102890)

rohan-varma committed 1 year ago

a748be93

[Composable] fully_shard load_optim test (#102692)

rohan-varma committed 1 year ago

5b623d6c

[FSDP/DDP] Unify _cast_forward_inputs (#102680)

rohan-varma committed 1 year ago

88ce6215

[FSDP/AC] checkpoint_wrapper acccept auto_wrap_policy (#102672)

rohan-varma committed 1 year ago

957ea485

[FSDP][Docs] Document get_state_dict_type (#102658)

rohan-varma committed 1 year ago

df40ec82

[FSDP] Document optim_state_dict_config in method (#102657)

rohan-varma committed 1 year ago

c6d0fe39

[FSDP] Use INFO instead of DETAIL for warning logs (#102639)

rohan-varma committed 1 year ago

beb7131c

[FSDP][ez] Type optimizer correctly (#102637)

rohan-varma committed 1 year ago

4d516f44

Log modules FSDP hooks fire for (#102508)

rohan-varma committed 1 year ago

e66c498d

[pt2] add `SymInt` support for `linalg.vander` (#102469)

nkaretnikov committed 1 year ago

757791d1

increase clang-tidy coverage to more c10 source files (#102902)

cyyever committed 1 year ago

87cbfe95

[vision hash update] update the pinned vision hash (#102919)

pytorchupdatebot committed 1 year ago

992bffe5

make device normalization more generic in faketensor (#102519)

heidongxianhua committed 1 year ago

9d20b47e

Add a new UNSTABLE category in trymerge (#102784)

huydhn committed 1 year ago

85efacee

Replace _dynamo.config with an object instead of module (#96455)

qihqi committed 1 year ago

3864207c

[Executorch][codegen] Add ETKernelIndex for aggregating all kernels for kernel (#102874)

larryliu0820 committed 1 year ago

eebe0ee1

Revert "[DTensor][3/N] add DTensor constructor function: full (#101436)"

pytorchmergebot committed 1 year ago

0f672e8c

Initialize optimizer in dynamo to avoid graph break and tracing slowness (#102640)

mlazos committed 1 year ago

c46af25b

[quant][pt2e][be] Remove _input_output_share_observers and _reuse_input_obs_or_fq from QuantizationAnnotation (#102854)

jerryzh168 committed 1 year ago

eb0971cf

Feature:To add --tolerance option to benchmark scripts (#102218)

vinayburugu committed 1 year ago

82154688

Introduce fast path for cuda_equal (#102714)

houseroad committed 1 year ago

12375022

[BE] Fix `lintrunner init` on python 3.11 (#102889)

ZainRizvi committed 1 year ago

4254e052

Fix optimizer cuda health check graph break (can be done in the compiler) (#102765)

mlazos committed 1 year ago

00f1bb09

[Dynamo] Fix if condition on UnspecializedNNModuleVariable (#102583)

yanboliang committed 1 year ago

d92bb036

Remove `check` from `_prims_common`, replace with `torch._check*` (#102219)

kurtamohler committed 1 year ago

a84bb270

[dynamo] test attaching attributes to an OptimizedModule (#102781)

davidberard98 committed 1 year ago

1035e33b

Newer Older