pytorch/xla

Pull Requests Commits

dlibenzi committed 6 years ago

ed35bab0

Add fast has generation from IR Node objects to be able to quickly detect graph changes.

dlibenzi committed 6 years ago

f4a51c0a

Create device shape out of IR Node shape.

dlibenzi committed 6 years ago

1ea6aad0

Add dereference operator to access IR Node pointer from NodeOperand, and added more testing for the IR Node replace operation.

dlibenzi committed 6 years ago

0447d50c

Added automatic conversion to IR Output for the IR NodeOperand.

dlibenzi committed 6 years ago

06df5fb4

Use the proper output index for applying operations.

dlibenzi committed 6 years ago

0dac568b

Turn IR operation to act on NodeOperand in order to support multi-output IR Node within XLA tensors.

dlibenzi committed 6 years ago

b9b8dd09

dlibenzi committed 6 years ago

144756a7

Added multi-ops IR Node return facility.

dlibenzi committed 6 years ago

6fc3d1d9

Added maxpool and log softmax backward tests.

dlibenzi committed 6 years ago

358f8b82

Wrap the generated regsitration function with an std::once_flag.

dlibenzi committed 6 years ago

1501692f

Added convolution backward test.

dlibenzi committed 6 years ago

cc5b447e

Correctly set sizes, strides and num elements.

dlibenzi committed 6 years ago

c04d640d

Account for undefined tensors coming out of ATEN results.

dlibenzi committed 6 years ago

e495909a

Drop the USE_XRT env config, as there is no more XLA option.

dlibenzi committed 6 years ago

00f33e16

Use at::Scalar instead of double for IR Scalar Node.

dlibenzi committed 6 years ago

855c30f6

Make C++ test utility more resilient to different inputs (XLA vs ATEN tensors, variables or raw tensors).

dlibenzi committed 6 years ago

535a1c2d

Move the variable creation of ToTensor() APIs out to the Python binding as otherwise autograd barks.

dlibenzi committed 6 years ago

1ee3a24e

Fix PT/XLA tensor shape generation to create tuple when input is tuple.

dlibenzi committed 6 years ago

0cbd52aa

Generate debug metric code for intercepted functions.

dlibenzi committed 6 years ago

1a7b7ee3

Remove dummy file as the folder has a README file now.

dlibenzi committed 6 years ago

9478c8a9

Address review comments

asuhan committed 6 years ago

e1d2be53

Add XLA lowerings to AtenXlaType

asuhan committed 6 years ago

d7781763

Fix XlaTensorDevice

asuhan committed 6 years ago

c795036a

Make scripts/gen.py use newly added enum values

asuhan committed 6 years ago

756e7d6c

The XLA TensorImpl needs an empty Storage, as other APIs in PT wants to access it to fetch things like device or data type.

dlibenzi committed 6 years ago

b5c2a57a

Add XLA tensor implementation method override, as this is used by make_variable() on XLA tensors, and if not overridden causes crashes trying to access not existent Storage.

dlibenzi committed 6 years ago

caa3cc79

Added a README and removed the now-applied PR patch.

dlibenzi committed 6 years ago

f3e974f9

Add infrastructure to apply outstanding PT PRs.

dlibenzi committed 6 years ago

a6a14550

Fix code generation for new IntArrayRef type.

dlibenzi committed 6 years ago

f86fb971

Older