pytorch/xla

Pull Requests Commits

Add full precision option to matmul

Alex Şuhan committed 7 years ago

472d4767

Fix aten::randint() infinite loop by routing it to CPU.

dlibenzi committed 7 years ago

419d05f2

Remove prim::None handler from translator

Alex Şuhan committed 7 years ago

4b62ea82

Hack nll_loss_forward in translator

Alex Şuhan committed 7 years ago

26c90ca9

Handle None constants in translator

Alex Şuhan committed 7 years ago

0b036ed3

Prepare for some of the optimizer's curved balls.

dlibenzi committed 7 years ago

ebcd365a

Made basic aruthmetic functions static.

dlibenzi committed 7 years ago

0a76b8be

Add scalar version of ATEN in place mul operation.

dlibenzi committed 7 years ago

6edfd85b

Alex Şuhan committed 7 years ago

434dc94e

Add at::ceil and at::floor

vishwakftw committed 7 years ago

e39be57f

Add rsub to ATen XLA tensor

asuhan committed 7 years ago

badbc523

Add leaky_relu to ATen XLA tensor

asuhan committed 7 years ago

deb7d34d

Added aten::matmul operation.

dlibenzi committed 7 years ago

86f71585

Added aten::cat operation.

dlibenzi committed 7 years ago

86f7e31f

Fix Permute::ToString

asuhan committed 7 years ago

7eae43e2

Enable the conversion API to work across allowed types.

dlibenzi committed 7 years ago

cb21840f

Make view() a static method.

dlibenzi committed 7 years ago

29f98ebf

Add text graph output.

dlibenzi committed 7 years ago

995935af

Add repeat to ATen XLA tensor

asuhan committed 7 years ago

7b65e220

Add at::atan, at::tan, at::tanh

vishwakftw committed 7 years ago

6d9bf207

Added support for exporting IR graphs in DOT format.

dlibenzi committed 7 years ago

830c6e50

Added aten::nll_loss_forward (called by Python MNIST test).

dlibenzi committed 7 years ago

f6dbb693

Made the NotSupported IR Node to have its own ID and carry a provenience description.

dlibenzi committed 7 years ago

a2eba815

Added _log_softmax() ATEN override (needed by MNIST).

dlibenzi committed 7 years ago

6024b8f7

Support post-order generation from multiple nodes.

dlibenzi committed 7 years ago

fedea843

Consolidate tensor type mapping code.

dlibenzi committed 7 years ago

e4548028

Generate conversions TPU can understand.

dlibenzi committed 7 years ago

16868b0b

Fix adaptive AVG pool and XLA where ops.

dlibenzi committed 7 years ago

13021473

Added cast operations.

dlibenzi committed 7 years ago

7356cde7

Add at::erf and at::erfc operators

vishwakftw committed 7 years ago

fc341bd0

Newer Older