strisunshinewentingwang/pytorch

Pull Requests Commits

[ROCm] reduce tolerance for triangular solve with well_conditioned set to True (#104425)

alugorey committed 1 year ago

c89aec20

[easy] Minor torch.load docs fix (#105876)

mikaylagawarecki committed 1 year ago

9c458942

add basic cuda support for float8 dtypes (#105807)

vkuzo committed 1 year ago

8b34fa5e

[PyTorch][ET] Collect Process Groups Mapping Info (#104373)

louisfeng committed 1 year ago

3a01c056

[ONNX] Export module as function (#105618)

BowenBao committed 1 year ago

00ee38c6

[ONNX] Improve shape inference for Slice (#105755)

ilyasher committed 1 year ago

ec337337

Support dynamic shapes in TritonTemplates (#105295)

ipiszy committed 1 year ago

98956c53

[Compiled Autograd] Move to torch::dynamo::autograd namespace (#105854)

jansel committed 1 year ago

26e3b402

Provide a refined upper bound for nonzero when original numel is static (#105843)

ezyang committed 1 year ago

5403c777

Slightly improve AOTAutograd logging with ViewAndMutationMeta (#105702)

ezyang committed 1 year ago

cc137342

Add meta registration for foreach_maximum_.List (#105864)

janeyx99 committed 1 year ago

5fec1f93

[FSDP][Docs] Tidy up FSDP ctor/api docs (#105847)

awgu committed 1 year ago

6655b652

[ONNX] Passes to reuse existing fake mode if possible (#105764)

BowenBao committed 1 year ago

65bce811

Cleanup calculate-docker-image (#105752)

huydhn committed 1 year ago

8047ce05

[inductor] triton_utils.config_of: check for divisibility by 16, even when expr is not an Integer (#105743)

davidberard98 committed 1 year ago

becb8dc9

[functional collective] parameterize allreduce tests (#105604)

wanchaol committed 1 year ago

8b942800

Meff Attn Bias (#104310)

drisspg committed 1 year ago

54535081

[exir] Add deepcopy to ExportedProgram (#105852)

JacobSzwejbka committed 1 year ago

9d62c5fa

Compiled autograd (#103822)

jansel committed 1 year ago

c902b84e

Remove unnecessary simplification in guard_lt (#105842)

ezyang committed 1 year ago

14304afd

Use hipsolver for default svd case on ROCm (#103540)

alugorey committed 1 year ago

b78341dd

Strengthen ConstantVariable invariants (#105796)

voznesenskym committed 1 year ago

bf693f20

Add a version to this signpost so I can tell if packages have taken updates (#105735)

ezyang committed 1 year ago

d2ee3d06

Implement NAdamW optimizer (#103881)

janEbert committed 1 year ago

b0708654

Add SparseCsrCPU and SparseCsrCUDA dispatch to sum.dim_IntList (#99292)

yanbing-j committed 1 year ago

a5404351

Use the newer g5.12xlarge instead of g3.16xlarge for multigpu tests (#105759)

huydhn committed 1 year ago

fb0ffeec

Tweak dynamic=False behavior (#105715)

ezyang committed 1 year ago

3045e84e

[BE] remove deprecated attributes from distributed_c10d (#105753)

H-Huang committed 1 year ago

0ab74044

use UBSAN on some tests (#103655)

cyyever committed 1 year ago

b8eb827d

[ROCm] Skip test_jit_cudnn_extension on ROCm (#105805)

pruthvistony committed 1 year ago

68dce237

Newer Older