strisunshinewentingwang/pytorch

Pull Requests Commits

[inductor] Create triton_helpers module for helper functions (#99880)

peterbell10 committed 1 year ago

ed2eb13d

[c10d] Scalable PG initiation. (#99931)

Rodrigo Kumpera committed 1 year ago

ad21890f

sparse.sum backward: short circuit on zero/empty grad (#98838)

nikitaved committed 1 year ago

2eab5abb

Add support for serializing real tensor data in after aot minifier (#99834)

ezyang committed 1 year ago

67e0913d

relu/threshold backward for sparse: enable 0-nnz grads (#98935)

nikitaved committed 1 year ago

5cfaea15

Change caffe2 branch links to main (#100129)

kit1980 committed 1 year ago

c2402a92

Include all mkl/mkldnn related test files to CPU ATen backend (#99592)

jgong5 committed 1 year ago

77a37a54

Basic dynamo support for traceable collectives (#94440)

wconstab committed 1 year ago

100a25d0

[CUDA] Switch to `at::empty` in `max_pool3d_with_indices_backward_cuda` (#100138)

eqy committed 1 year ago

925a3788

Making fsdp device-agnostic for custom-backend which implement cuda-semantics (#99024)

medivh-xp committed 1 year ago

859e82a7

[inductor] fix _print_Pow given reciprocal of dynamic dim with float exponent (#100090)

jgong5 committed 1 year ago

4456e932

inductor: enhance conv+binary fusion path test for cpu path (#100058)

XiaobingSuper committed 1 year ago

569eff85

fix missing-prototypes warnings in torch_cpu (Part 1) (#100053)

cyyever committed 1 year ago

e2480164

[dynamo] Add ddp_graphs artifact (#100021)

bertmaher committed 1 year ago

e0bf51d3

[inductor] logger message fix in split_cat (#100088)

jgong5 committed 1 year ago

1504bdf9

Rename DispatchKey.PrivateUse1 to custom device in torchgen. (#99406)

huihoaan committed 1 year ago

c0ecd989

inductor: simplify the test_mkldnn_pattern_matcher.py code (#100057)

XiaobingSuper committed 1 year ago

3588688a

[ONNX] Fix type annotation for 'fx_to_onnxscript' (#100050)

BowenBao committed 1 year ago

13259fe8

Refactors/improvements in _inductor/fx_passes (#100063)

jansel committed 1 year ago

3f5d768b

[Tensor Parallel] Simplify distribute for MHA (#100046)

rohan-varma committed 1 year ago

be8c7c06

add sm80orlater check for bfloat test in test_torchinductor (#98034)

Fuzzkatt committed 1 year ago

97f4af3f

Don't apply _Py_OPCODE twice (#97986)

AlekseiNikiforovIBM committed 1 year ago

bc0c74bc

Introduce FXGraphExtractor into torch.onnx.dynamo_export (#99940)

Thiago Crepaldi committed 1 year ago

32a67e42

[inductor] fix inconsistent behaviours when padding size is zero (#100082)

cdzhan committed 1 year ago

763e0a90

[BE][DTensor] add DeviceMesh test to periodic testing list (#100029)

XilunWu committed 1 year ago

19e81b7b

Fix prims unbind if given dimension size is 0 (#100122)

yanboliang committed 1 year ago

4c6f7cbc

[Dynamo] Fix constructing lazy submodule inside of lazy module's initialize_parameters (#100047)

yanboliang committed 1 year ago

2989d6c9

enable -Werror=sign-compare in our Bazel build (#98671)

mikey dagitses committed 1 year ago

fab2e397

[dynamo] Make bytecode logging off-by-default (#100093)

bertmaher committed 1 year ago

6789342a

Add a new hook (#99854)

openrichardfb committed 1 year ago

c523d7d8

Newer Older