strisunshinewentingwang/pytorch

Pull Requests Commits

trivially convert all memcpy destinations to mutable_data_ptr (#98753)

mikey dagitses committed 1 year ago

3a510e39

Respect sharded dimensions when aten expaned/view consumes SymInt values (#99058)

mrshenli committed 1 year ago

40aaacd4

[torch package][easy] Make all the save/load tests use buffers (#98798)

kunalb committed 1 year ago

d365d9ed

[MPS] Fix macOS build (#99139)

qqaatw committed 1 year ago

21035462

make ATen/native/cuda/group_norm_kernel.cu data_ptr-correct (#99041)

mikey dagitses committed 1 year ago

05a55b96

Add vmap support for `torch.nn.functional.smooth_l1_loss` (#98357)

yhl48 committed 1 year ago

298cc5c6

Make summarize_perf.py work with perf-compare (#99095)

lezcano committed 1 year ago

1e78a2ed

Don't specialize when indexing by SymInt (#99123)

ezyang committed 1 year ago

ca735ac8

Put GraphArg on the node meta. (#99068)

ezyang committed 1 year ago

0963e118

Expandable blocks in allocator (#96995)

zdevito committed 1 year ago

6a50b83b

[MPS] Add ASSERT_ONLY_METHOD_OPERATORS Finish (#99021)

qqaatw committed 1 year ago

2494e625

[MPS] Add ASSERT_ONLY_METHOD_OPERATORS Part 5 (#99020)

qqaatw committed 1 year ago

7ddeb8d3

[MPS] Add ASSERT_ONLY_METHOD_OPERATORS Part 4 (#99019)

qqaatw committed 1 year ago

70120f2f

[MPS] Add ASSERT_ONLY_METHOD_OPERATORS Part 3 (#99018)

qqaatw committed 1 year ago

f043ff2c

[MPS] Add ASSERT_ONLY_METHOD_OPERATORS Part 2 (#99017)

qqaatw committed 1 year ago

be50c1c4

[MPS] Add ASSERT_ONLY_METHOD_OPERATORS Part 1 (#99016)

qqaatw committed 1 year ago

4ddaab84

make native/cuda/EmbeddingBackwardKernel.cu data_ptr-correct (#99027)

mikey dagitses committed 1 year ago

27049f3f

[SPMD] Upstream partial_lower (#99069)

fegin committed 1 year ago

bce73088

Re-PR of 90269 - Force all nn_module associated tensors to be static (#99108)

voznesenskym committed 1 year ago

10fbdcf7

[spmd][easy] delete unused optim states during compile (#99061)

wanchaol committed 1 year ago

b3e63baf

[dtensor] redistributed by default take self mesh instead (#99060)

wanchaol committed 1 year ago

55a1dc7f

inductor: fix timeout in ModularIndexing (#98841)

blzheng committed 1 year ago

cdef4f07

[torch.compile] Add sympy.core.relational.Relational to inductor.ir (#98971)

YJ Shi committed 1 year ago

0a7baaba

[ONNX] Remove duplicated code from previous rebase (#99072)

BowenBao committed 1 year ago

9d62f771

GraphArg is always length one, adjust APIs accordingly (#99059)

ezyang committed 1 year ago

cd078d37

Remove dead wrap_sym (#99049)

ezyang committed 1 year ago

e613a419

Delete dead orig_graphargs (#99047)

ezyang committed 1 year ago

21ed07ce

Change unspec ints to not be duck-sized (#99010)

ezyang committed 1 year ago

cc345d18

make ATen/native/cuda/Loss.cu data_ptr-correct (#99073)

mikey dagitses committed 1 year ago

a6a3df08

make ATen/native/cuda/MultiTensorApply.cuh data_ptr-correct (#99081)

mikey dagitses committed 1 year ago

8382e91b

Newer Older