pytorch/xla

Pull Requests Commits

Use two-pass lowerings for argmin and argmax

asuhan committed 7 years ago

c9d654ad

Fix pending PR extension

asuhan committed 7 years ago

a5d996d0

Fix XlaModule constructor after PyTorch changes

asuhan committed 7 years ago

470b15e1

Local fix for max_pool2d differentiation

asuhan committed 7 years ago

160d9cb5

Use xla::TorchGather to lower Gather

asuhan committed 7 years ago

eb769a38

Use xla::TorchIndexSelect to lower IndexSelect

asuhan committed 7 years ago

1080e7f7

Add pending PyTorch patch fixing equal for boolean tensors

asuhan committed 7 years ago

bde41239

Update to latest TF

asuhan committed 7 years ago

73919342

Add step-mark operation, and split sync-tensors into a separate operation.

dlibenzi committed 7 years ago

9e59aa5c

Simplify XLA tensor code generator further, after the recent ATEN changes.

dlibenzi committed 7 years ago

de703fa7

Fix XLA tensor code generation to cope with the PyTorch/ATEN changes.

dlibenzi committed 7 years ago

2ba3696c

Throw when trying to access storage of an XLA tensor

asuhan committed 7 years ago

0ec9b5ac

Bump up the IR graph trim thresholds.

dlibenzi committed 7 years ago

2bde0d0f

Extend the graph-saving debug utility to more places where we enter the XLA tensor world for execution.

dlibenzi committed 7 years ago

25d099bf

Read async status in wait operation.

dlibenzi committed 7 years ago

f3221842

Remove Hash() default to DataHash(), and restrict to arithmetic for it.

dlibenzi committed 7 years ago

b2f6d951

Fix code by coping with the fact that ATEN copy() is no more.

dlibenzi committed 7 years ago

5dd0ea12

Return a torch.device consistently.

dlibenzi committed 7 years ago

d2e239b0

Fix index_put starting with null indices

asuhan committed 7 years ago

64de8e6a

Add base support for ATEN Bool tensor type.

dlibenzi committed 7 years ago

04faa305

Send tensors to XLA device before running tests, in case of tensor-modifying test functions.

dlibenzi committed 7 years ago

9199f36f

Made aten::clamp operation to use possibly cache min/max arguments, w/out generating ops::Scalar.

dlibenzi committed 7 years ago

82e82f5a

Fix multi-index operation which starts with null indices

asuhan committed 7 years ago

0c543068

Make unsqueeze() return a view.

dlibenzi committed 7 years ago

5b6085ce

Promote sides of bitwise logical operations

asuhan committed 7 years ago

efd20643

Add randperm to ATen XLA tensor

asuhan committed 7 years ago

002eb2d3

Fixed zero-sized slice operation.

dlibenzi committed 7 years ago

695c93f7

Fix unsqueeze operation symbol.

dlibenzi committed 7 years ago

1d95c624

Add s__th_eq wrapper to ATen XLA tensor

asuhan committed 7 years ago

04fd6ddd

Route more scalars to (possibly) device data.

dlibenzi committed 7 years ago

3086d260

Newer Older