pytorch/xla

Pull Requests Commits

Added c10 linkage.

dlibenzi committed 6 years ago

55ecb142

Fix bolted in path.

dlibenzi committed 6 years ago

fd50735b

Do not assume library folder.

dlibenzi committed 6 years ago

c976167c

Few more tweaks and added run shell script.

dlibenzi committed 6 years ago

8acb6f51

Seems to be working now.

dlibenzi committed 6 years ago

dfbaf1a1

Added simple test.

dlibenzi committed 6 years ago

7750f962

Trying to get c++ unittests working.

dlibenzi committed 6 years ago

ec221e0f

Moved arithmetic IR ops.

dlibenzi committed 6 years ago

11efb7a5

Enable XLA tensor data to persist.

dlibenzi committed 6 years ago

cc20fdf0

dlibenzi committed 6 years ago

74703986

Allow easier creation of arithmetic ops.

dlibenzi committed 6 years ago

1baee19d

Added utility APIs to create IR Node instead of replicating make_shared.

dlibenzi committed 6 years ago

43f0c84c

Fix review comments.

dlibenzi committed 6 years ago

04032a51

Change the IR node set API name.

dlibenzi committed 6 years ago

88e5455e

Bring back graph size.

dlibenzi committed 6 years ago

b37a0e7a

Added more comments and changed the lowering context constructor.

dlibenzi committed 6 years ago

513a5619

Revised the ToString implementations.

dlibenzi committed 6 years ago

92a91bc3

Builds, test pass, still need to figure out the graph size.

dlibenzi committed 6 years ago

fec3dbe8

Removed old Graph context.

dlibenzi committed 6 years ago

558a089a

Moving forward with XLA tensor integration.

dlibenzi committed 6 years ago

3790b8eb

Add Shape to IR Node and implement generic Node.

dlibenzi committed 6 years ago

ade157d7

Squash all warnings

asuhan committed 6 years ago

6da572fa

Update the PyTorch commit id to latest

asuhan committed 6 years ago

7c353931

Drooped util API and initializing the vector in spite of narrowing.

dlibenzi committed 6 years ago

7722d4f8

Use the equivalent Tensor shape for the input tensor. Added hack comment.

dlibenzi committed 6 years ago

aae5e4c2

Use the new tensor based XRT allocation.

dlibenzi committed 6 years ago

949bbb01

Revise weird XlaOps() accessory.

dlibenzi committed 6 years ago

009843ef

Move to new TF HEAD and fetch new XRT bits.

dlibenzi committed 6 years ago

3dd0e6ae

Add XLA data leaf node to the IR.

dlibenzi committed 6 years ago

923bb26b

Added ops::Constant node.

dlibenzi committed 6 years ago

28580a9f

Older