pytorch/xla

Pull Requests Commits

Made autogen handle cases similar to copy(), where a different "reference param" needs to be picked for device parameter.

dlibenzi committed 7 years ago

eceb8d87

Handle copy() operation directly.

dlibenzi committed 7 years ago

1904677d

Remove a core PyTorch patch which landed

Alex Şuhan committed 7 years ago

00c9919e

Alex Şuhan committed 7 years ago

758d1106

Remove pun-to-CPU for aten::topk not sorted.

dlibenzi committed 7 years ago

0a3846e6

Normalize positions and dimensions for certain operators.

dlibenzi committed 7 years ago

047620d5

Added aten::topk operation.

dlibenzi committed 7 years ago

a2581552

Added aten::kthvalue() operation.

dlibenzi committed 7 years ago

80722f3e

Add at::prod family of operations

vishwakftw committed 7 years ago

9978d495

Move GetCanonicalDimensionIndex to XlaHelpers

asuhan committed 7 years ago

637d0270

Add unbind to ATen XLA tensor

asuhan committed 7 years ago

9b8c33dc

Canonicalize select index and add C++ test

asuhan committed 7 years ago

670656a8

Added aten::log10 operation and restructured aten::log2 to use the same LogBase API.

dlibenzi committed 7 years ago

6bd0b1a3

Added aten::split_with_sizes operation.

dlibenzi committed 7 years ago

b515eb1c

Add at::erfinv operator

vishwakftw committed 7 years ago

7457d841

Added aten::meshgrid operation. Fix expand operation.

dlibenzi committed 7 years ago

156a7056

Add broadcast_tensors to ATen XLA tensor

asuhan committed 7 years ago

626ccd6e

Added aten::arange family oprations.

dlibenzi committed 7 years ago

2483c748

Added aten::embedding operation routed via native:: support.

dlibenzi committed 7 years ago

965abdac

Added aten::dropout operation.

dlibenzi committed 7 years ago

f8864d78

vishwakftw committed 7 years ago

d31edcbe

Add at::sign operator

vishwakftw committed 7 years ago

46d80265

Added aten::log2 operation.

dlibenzi committed 7 years ago

3d165ac4

Added inplace aten::relu_ operation.

dlibenzi committed 7 years ago

00a76e70

Handle (noop) the aten::contiguous operation.

dlibenzi committed 7 years ago

fe31b754

Added aten::any and aten::all operations.

dlibenzi committed 7 years ago

874ac810

Add more complex einsum tests

Alex Şuhan committed 7 years ago

dae5b02b

Add reshape to ATen XLA tensor

Alex Şuhan committed 7 years ago

77656686

Complete support for einsum via PyTorch lowering

Alex Şuhan committed 7 years ago

c54fb2a8

Added aten::index_select operation.

dlibenzi committed 7 years ago

6ca69874

Older