pytorch/xla

Pull Requests Commits

Add leaky_relu_backward to ATen XLA tensor

asuhan committed 6 years ago

4e008bb8

Replace remaining const ref at::Scalar

asuhan committed 6 years ago

709fe190

Add elu_backward to ATen XLA tensor

asuhan committed 6 years ago

e287c7b1

Added timed execution in DEBUG=1 mode to trace performance.

dlibenzi committed 6 years ago

8e84bee6

Add linear to ATen XLA tensor

asuhan committed 6 years ago

3bb45d68

Fix the fixup script to cope with some other cases of Conda site-packages structures.

dlibenzi committed 6 years ago

a9646f81

Move the XLA tensor methods implementing the ATEN function into a new tensor_methods.cpp file.

dlibenzi committed 6 years ago

4e662dde

Replication devices should return xla:* devices. Added get-count API for it as well.

dlibenzi committed 6 years ago

93e9f8b9

Add remainder to ATen XLA tensor

asuhan committed 6 years ago

fd4e6d1c

Adjust the scalar type of cumsum, cumprod

asuhan committed 6 years ago

34075c4d

Add cumsum and cumprod to ATen XLA tensor

asuhan committed 6 years ago

52c856ee

Scale gradients CRS to inverse of number of replicas.

dlibenzi committed 6 years ago

07f06bf1

Add narrow_copy to ATen XLA tensor

asuhan committed 6 years ago

1cc7e828

Refresh the autograd PR patch.

dlibenzi committed 6 years ago

02170d6a

Fix typo leftout by previous PR.

dlibenzi committed 6 years ago

70dfa105

Add chain_matmul to ATen XLA tensor

asuhan committed 6 years ago

09a29a3d

Propagate replication devices into XRT compile instances.

dlibenzi committed 6 years ago

eedd61ae

Prepare for setting up Nx replication mode.

dlibenzi committed 6 years ago

a61fe633

Update XLA/autograd PR patch.

dlibenzi committed 6 years ago

92b7ddf1

Added aten::clone operation.

dlibenzi committed 6 years ago

31d31fb6

Pass vector by value instead of slice to ops

asuhan committed 6 years ago

3b1e368e

Add shift operations to ATen XLA tensor

asuhan committed 6 years ago

1b4e9668

Clear the view once we set the XLA data. Added Resnet18 parallel test as well.

dlibenzi committed 6 years ago

7ea5cb99

Leave indices empty for permutation ViewInfo

asuhan committed 6 years ago

9aee1c4a

Add simple cases of as_strided and as_strided_ to ATen XLA tensor

asuhan committed 6 years ago

fe3ad982

Add in-place bitwise operations to ATen XLA tensor

asuhan committed 6 years ago

5a6c8c8e

Do not explicitly track graph size but lazily evaluate it and possibly trim.

dlibenzi committed 6 years ago

411f15a5

Added aten::threshold_ operation.

dlibenzi committed 6 years ago

63db8794

Uniformly pass at::Scalar by value

asuhan committed 6 years ago

bc11fc0b

Fix tensor.to() API.

dlibenzi committed 6 years ago

184fe590

Newer Older