strisunshinewentingwang/pytorch

Pull Requests Commits

inductor: rewrite mkldnn fx fusion using pattern_matcher(binary) (#97141)

XiaobingSuper committed 1 year ago

9c98f2ce

inductor: rewrite mkldnn fx fusion using pattern_matcher(conv_transpose_unary) (#97140)

XiaobingSuper committed 1 year ago

d3a1a772

inductor: fix the issue of cat missing dim argument for sink_cat_after_pointwise (#98901)

XiaobingSuper committed 1 year ago

73c3cb71

inductor: rewrite mkldnn fx fusion using pattern_matcher(linear_unary) (#97139)

XiaobingSuper committed 1 year ago

562e5d49

inductor: rewrite mkldnn fx fusion using pattern_matcher(conv_unary) (#97007)

XiaobingSuper committed 1 year ago

c214c503

[BE] Use `Literal` from `typing` (#98846)

malfet committed 1 year ago

0be65069

[MPS] Expose mps package in torch (#98837)

qqaatw committed 1 year ago

6ff32b55

Skip dtensor ops on CPU-only runner due to flaky timeout (#98868)

huydhn committed 1 year ago

d3a35956

[Gloo][BE] Print stacktrace on collectFullMesh (#98810)

rohan-varma committed 1 year ago

60ebb2f1

Add Symbool support in python to C++ translation (#98453)

tugsbayasgalan committed 1 year ago

39fd7f94

torch.compile benchmark utility (#97699)

msaroufim committed 1 year ago

bc8cb62b

Enable fake_crossref unit tests on rocm (#97368)

BLOrange-AMD committed 1 year ago

455795c7

[BE] Move mobile builds to python-3.8 (#98886)

malfet committed 1 year ago

9c5473b7

[export] Constraints API (#98433)

angelayi committed 1 year ago

1510eb40

[llvm-17][ORC] Fix for move most ORC APIs to ExecutorAddr, introduce ExecutorSymbolDef. (#98811)

ayermolo committed 1 year ago

ac5025cd

[SPMD] Introduce remove_copy_for_optimizer optimization (#98580)

fegin committed 1 year ago

f3080997

[SPMD] Add optimizer states and steps to the return (#98579)

fegin committed 1 year ago

40132069

[SPMD] Introduce schedule_comm_wait (#98578)

fegin committed 1 year ago

07a1378f

[SPMD] Introduce graph_optimization_pass and comm_fusion_with_cat (#98285)

fegin committed 1 year ago

dd3e2ddc

[nccl] Remove lock for nccl collective launch for 2.0+ (#97904)

xw285cornell committed 1 year ago

78ad800a

[memory viz] support larger visualizations (#98865)

zdevito committed 1 year ago

e37986d4

[dynamo] Fix bug with torch._dynamo.skip (#98862)

anijain2305 committed 1 year ago

a2e0f512

[SPMD] Add a dump_graphs_to_files utils to facilitate graph transformation debug (#98284)

fegin committed 1 year ago

2de67eaa

Fix 'fully_shard' may determine compute device incorrectly (#98831)

medivh-xp committed 1 year ago

09621148

[Easy] Reuse `source` variable in `wrap_tensor` (#98845)

awgu committed 1 year ago

c93ff384

[ONNX] Skip flaky dynamic tests before ORT==1.15 in fx exporter (#98856)

titaiwangms committed 1 year ago

ad373efe

Fix Lint (#98873)

PaliC committed 1 year ago

6cbe5c5e

[MTPG] add all_to_all collective to MTPG (#98791)

XilunWu committed 1 year ago

89894115

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

qihqi committed 1 year ago

420104a8

[SPMD] Add the default graph module transformation that is applied after tracing and expansion (#98182)

fegin committed 1 year ago

06c206ce

Newer Older