strisunshinewentingwang/pytorch

Pull Requests Commits

[FSDP][3/N]_shard_utils update for dtensor state_dict support (#103479)

wz337 committed 1 year ago

d991ce6d

[CI] Enable inductor dynamic accuracy test on cpu device (#103387)

chuanqi129 committed 1 year ago

3c5ac4ba

[Dynamo] Fix lineinfo generation on PY3.11+ (#103525)

malfet committed 1 year ago

f0832914

[vision hash update] update the pinned vision hash (#103560)

pytorchupdatebot committed 1 year ago

193d8412

inductor: using int64 as index dtype for slice_scatter (#103511)

XiaobingSuper committed 1 year ago

674d18c1

[inductor] fix benchmark call for inplace update (#103547)

shunting314 committed 1 year ago

2e1369d7

`default` should be used as default value in `boolean_dispatch` (#103463)

lkct committed 1 year ago

87616198

[Pytorch] aten::zero_ (#103042)

lucylq committed 1 year ago

cbea85b4

Update lr_scheduler.py to check the type of eta_min (#97003)

ravikiranrao committed 1 year ago

83407622

Refactor tests for dynamic shapes (#103542)

ezyang committed 1 year ago

2f5fef59

extend serialization for tensor metadata (#99808)

fakeYan committed 1 year ago

b7777c81

Using dynamic allocation buffer and dynamic threads on scan with index (#103502)

mfkasim1 committed 1 year ago

ce0a5119

Make DDPOptimizer handle subgraphs without outputs (#103488)

wconstab committed 1 year ago

fee01640

Use CUDA DSA in ATen (#95300)

r-barnes committed 1 year ago

93b0410e

Checking for nullptr in get_model_bytecode_version (#97149)

mmigdal-nv committed 1 year ago

6cc0f1c2

[reland][quant][pt2e] Annotate GRU module (#103358) (#103526)

jerryzh168 committed 1 year ago

0cd155b0

NCCL process group: avoid workEnqueue when capturing cuda graph (#103503)

bottler committed 1 year ago

02548800

Fix freezing tests (#103531)

eellison committed 1 year ago

25b6b95b

Strengthen partially supported invariant of base for chained sources (#103445)

voznesenskym committed 1 year ago

056bf951

Add type hint for retains_grad (#103528)

ringohoffman committed 1 year ago

bc2caa7f

[pt2] add `SymInt` support for `cosine_similarity` (#103400)

nkaretnikov committed 1 year ago

d38b651d

[pt2] add `SymInt` support for `bilinear` (#103396)

nkaretnikov committed 1 year ago

c0763443

[pt2] add metas for `avg_pool3d` and `avg_pool3d_backward` (#103392)

nkaretnikov committed 1 year ago

4a76fb49

[export] Serialize symbolic values (#103273)

angelayi committed 1 year ago

8dc60010

[ONNX] Add constant folding for Softmax op (#102861)

ilyasher committed 1 year ago

876695d4

Always register SHAPE_ENV guard (#103521)

ezyang committed 1 year ago

3804eb10

torch.compiler public namespace (#102182)

msaroufim committed 1 year ago

ea384cd3

Always apply new_empty special case in Dynamo (#103378)

ezyang committed 1 year ago

3596a853

[FSDP][2/n] add use_dtensor flag to both StateDictConfig and OptimStateDictConfig (#103477)

wz337 committed 1 year ago

51d21ffd

Unified aa_filter and aa_filter_075 for bicubic upsampling (#103510)

vfdev-5 committed 1 year ago

72931759

Newer Older