pytorch/xla

Pull Requests Commits

Generate the code before globbing.

dlibenzi committed 6 years ago

31e0805d

Document "things".

dlibenzi committed 6 years ago

0e52d3c8

dlibenzi committed 6 years ago

30a7f591

dlibenzi committed 6 years ago

5a2d9249

Install the Lark parser in the circleci build script.

dlibenzi committed 6 years ago

d7914ad3

Do not commit autogenerated files and add step to build process.

dlibenzi committed 6 years ago

d17ab0cc

Generate Type class code.

dlibenzi committed 6 years ago

0baa095a

Add missing include path for C++ test

asuhan committed 6 years ago

382df7f8

Handle infinite recursion when handling ATEN APIs which have no Tensor context to redirect to.

dlibenzi committed 6 years ago

6a8fa050

Fixes for XLA to ATEN generator script

asuhan committed 6 years ago

cb99065d

Generating single type instances in the class generation mode.

dlibenzi committed 6 years ago

24f979e1

Rewrite types to avoid using statements.

dlibenzi committed 6 years ago

9ad726a7

Generate ATEN Type class code.

dlibenzi committed 6 years ago

a4419284

Use tensor shape as output shape when XLA compiling.

dlibenzi committed 6 years ago

b8a18531

Generate proper TPU friendly shape in case of 4D/TPU tensors.

dlibenzi committed 6 years ago

2f847732

Fix review comments.

dlibenzi committed 6 years ago

88a3fe7c

Default to TPU device.

dlibenzi committed 6 years ago

86f5f8c1

Made the generated code compile.

dlibenzi committed 6 years ago

b9837556

Fix function name resolution.

dlibenzi committed 6 years ago

caa2fba1

Added XLA TensorImpl and created bridge APIs.

dlibenzi committed 6 years ago

50f3caa1

More refinements to the code generation (includes, namespaces,...).

dlibenzi committed 6 years ago

d1ab9c8e

Fix handling of void return functions.

dlibenzi committed 6 years ago

5854c733

Find the reference parameter instead of assuming 'self'.

dlibenzi committed 6 years ago

785b53db

Added at:: namespace.

dlibenzi committed 6 years ago

862ce178

More comments and refinements.

dlibenzi committed 6 years ago

4622151f

Blacklist CUDA functions and handle std::vector<Tensor> returns.

dlibenzi committed 6 years ago

702a2225

Full code generation with registration included.

dlibenzi committed 6 years ago

aed346a4

Refining more the generation code.

dlibenzi committed 6 years ago

0c4905f3

Change gen file exec bit.

dlibenzi committed 6 years ago

cee2e82c

Added first sketch of autogenerator for XLA to ATEN tensor plumbing.

dlibenzi committed 6 years ago

5e061523

Older