strisunshinewentingwang/pytorch

Pull Requests Commits

[pt2] add Python metas for `special` ops (#106683)

nkaretnikov committed 1 year ago

e7a3fb13

[TGIF][Inplace][Perf] Copy tensor to device with pinned memory & move copy weight sleep to getRecord (#106849)

sayitmemory committed 1 year ago

b897c57d

[executorch] Let custom ops registration code only import ATen headers (#107064)

larryliu0820 committed 1 year ago

ddd2f682

Keep fx node name consistent with aot_export (#107068)

gmagogsfm committed 1 year ago

f26aa2dc

[inductor] Optimize away zero-element loads (#107074)

peterbell10 committed 1 year ago

8472c24e

Add gfx90a target for ROCm CI (#106879)

jithunnair-amd committed 1 year ago

aa36e16f

[inductor][easy] add a missing parenthesis (#107001)

shunting314 committed 1 year ago

6f833821

Install torchrec/fbgemm from source in CI (#106808)

ezyang committed 1 year ago

5b04e9b6

Revert "Reordering tests experiment (#106347)"

pytorchmergebot committed 1 year ago

9858edd9

[device_mesh] move remaining collectives to a separate file (#107012)

wanchaol committed 1 year ago

c9cbcb24

[ONNX] Migrate to PT2 logging (#106592)

BowenBao committed 1 year ago

22095acf

Make the __call__ op of ExportedProgram follow calling convention. (#106186)

qihqi committed 1 year ago

5d09e499

[Dynamo x FSDP][2/x] Small changes to distributed to make it dynamo friendly (#106886)

voznesenskym committed 1 year ago

42660015

[inductor] graph replayer (#106952)

shunting314 committed 1 year ago

91778ada

[inductur][easy] show kernel name in str(ExternKernel) (#106990)

shunting314 committed 1 year ago

6730d5f9

Decomposition of bmm and mm for dot product (#106593)

mikekgfb committed 1 year ago

2c8f2482

Revert "Remove `set_default_dtype` from nn tests (#105775)"

pytorchmergebot committed 1 year ago

ec0f3fda

[inductor] fix test_dim_function_empty (#106994)

masnesral committed 1 year ago

3d00170b

[export] Support preserving calling convention to some modules. (#106798)

zhxchen17 committed 1 year ago

547ccae0

Revert "Add `_foreach_clamp` (#106574)"

pytorchmergebot committed 1 year ago

354484ea

Remove ExclusivelyOwned from register_dispatch_key (#106791)

Anstow committed 1 year ago

c9cdcb29

Propose nkaretnikov as general PrimTorch/meta/decomp reviewer (#106970)

ezyang committed 1 year ago

d97b18d7

[Dynamo] Improve PT2 fbcode logging observability (#106932)

yanboliang committed 1 year ago

fbfb9a16

Mark test_lstm_packed as slow (#107048)

clee2000 committed 1 year ago

1cfe2920

Add `isFloat8Type` predicate (#106977)

malfet committed 1 year ago

22a20d08

AsyncCollectiveTensor: dont sync on view ops (#105240)

wanchaol committed 1 year ago

5c48ff20

Implement decomposition for aten.rrelu_with_noise (#106812)

masnesral committed 1 year ago

e1659388

Skip Triton templates in MM max autotune with zero-size inputs (#106865)

aakhundov committed 1 year ago

b1b3f61f

Add multiprocess option to dynamo benchmarks (#106394)

H-Huang committed 1 year ago

656412f0

Fix test_cond_functionalized_aot_func_check_functional (#106889)

ydwu4 committed 1 year ago

3fe1dba0

Newer Older