pytorch/xla

Pull Requests Commits

Added aten::meshgrid operation. Fix expand operation.

dlibenzi committed 6 years ago

156a7056

Add broadcast_tensors to ATen XLA tensor

asuhan committed 6 years ago

626ccd6e

Added aten::arange family oprations.

dlibenzi committed 6 years ago

2483c748

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

dlibenzi committed 6 years ago

965abdac

Added aten::dropout operation.

dlibenzi committed 6 years ago

f8864d78

vishwakftw committed 6 years ago

d31edcbe

Add at::sign operator

vishwakftw committed 6 years ago

46d80265

Added aten::log2 operation.

dlibenzi committed 6 years ago

3d165ac4

Added inplace aten::relu_ operation.

dlibenzi committed 6 years ago

00a76e70

Handle (noop) the aten::contiguous operation.

dlibenzi committed 6 years ago

fe31b754

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

dlibenzi committed 6 years ago

874ac810

Add more complex einsum tests

asuhan committed 6 years ago

dae5b02b

Add reshape to ATen XLA tensor

asuhan committed 6 years ago

77656686

Complete support for einsum via PyTorch lowering

asuhan committed 6 years ago

c54fb2a8

Added aten::index_select operation.

dlibenzi committed 6 years ago

6ca69874

Add bmm to ATen XLA tensor

asuhan committed 6 years ago

782e0673

Add aten::full family and simplify code.

dlibenzi committed 6 years ago

fe0c7004

Change int in XLATensor signatures to xla::int64

asuhan committed 6 years ago

e4afba3f

Add in-place squeeze to ATen XLA tensor

asuhan committed 6 years ago

acedff08

Make parallel build the default.

dlibenzi committed 6 years ago

32ea2d79

Make Build{Triu, Tril} take xla::int64 diagonal

asuhan committed 6 years ago

64ea84d9

Add diagonal to ATen XLA tensor

asuhan committed 6 years ago

2755e2cc

Added end-to-end MNIST tensor training test.

dlibenzi committed 6 years ago

5dbac13d

Add pending PR patch which fixes the Python device property

asuhan committed 6 years ago

c99078fc

Be more accepting from things coming from optimizers into in-place operations.

dlibenzi committed 6 years ago

46e829e7

Handle empty*() and s_copy_() on XLA side.

dlibenzi committed 6 years ago

a98910a0

Add partial support for einsum to ATen XLA tensor

asuhan committed 6 years ago

ab15018b

The mul() operation as well might be coming with double ATEN tensor.

dlibenzi committed 6 years ago

83bd49e1

Fill in double values when lowering a scalar.

dlibenzi committed 6 years ago

a06cbc3f

Relax tolerance for TestMatmul_2x2

asuhan committed 6 years ago

cf2ddf0f

Newer Older