pytorch/xla

Pull Requests Commits

Fix diag on rank 1

asuhan committed 6 years ago

ae68a9b2

Fix the shape of the diagonal returned by SVD

asuhan committed 6 years ago

b931a6d2

Fix where implementation

asuhan committed 6 years ago

5059ebc8

Overlap XLA tensor sync with future XLA tensor work.

dlibenzi committed 6 years ago

10b8e428

Fix linear and softplus test tolerance on TPU

asuhan committed 6 years ago

2911d612

Finish up the ordering and do the XLA ATEN tensor sources as well.

dlibenzi committed 6 years ago

8097525c

Add argsort to ATen XLA tensor

asuhan committed 6 years ago

e5874ba3

Add one_hot to ATen XLA tensor

asuhan committed 6 years ago

ca7297d3

Add min, max (without indices) to ATen XLA tensor

asuhan committed 6 years ago

143d89d6

Adjust to removal of elementSizeInBytes in core PyTorch

asuhan committed 6 years ago

c0ec8392

Add {hard, soft}shrink[_backward] to ATen XLA tensor

asuhan committed 6 years ago

c3b32336

Add nuclear_norm to ATen XLA tensors

asuhan committed 6 years ago

0846a6e3

Add tanh_backward to ATen XLA tensor

asuhan committed 6 years ago

6b40a372

Add log_sigmoid_{backward} to ATen XLA tensor

asuhan committed 6 years ago

adb2a2aa

Add sigmoid_backward to ATen XLA tensor

asuhan committed 6 years ago

74fc337a

Sync to device data when sending tensors to XLA within the parallel loader threads.

dlibenzi committed 6 years ago

860b91fb

Add softplus_backward to ATen XLA tensor

asuhan committed 6 years ago

13736732

Add softplus to ATen XLA tensor

asuhan committed 6 years ago

64a2e4e4

Refresh autograd patch.

dlibenzi committed 6 years ago

877d2e41

Move softmax node creation to ops

asuhan committed 6 years ago

378b2044

Add softmax_backward to ATen XLA tensor

asuhan committed 6 years ago

fd0ec688

Fix data parallel to uniformily distribute over device queues.

dlibenzi committed 6 years ago

d476cf67

Remove redundant hardtanh and hardtanh_ from XLATensor

asuhan committed 6 years ago

38e9d966

Add hardtanh_ to ATen XLA tensor

asuhan committed 6 years ago

429277d2

Add hardtanh_backward to ATen XLA tensor

asuhan committed 6 years ago

baafb927

Add hardtanh to ATen XLA tensor

asuhan committed 6 years ago

7772e0b5

Cached compiled XLA computation during the tensors sync operatins.

dlibenzi committed 6 years ago

2d5a5589

Refactor lowering utils to move post-order lowering into a spearate function. Make single argument ctor explicit as well.

dlibenzi committed 6 years ago

3d488c88

Add counter for IR graph trimming and use the new XLA builder API to get the full trace.

dlibenzi committed 6 years ago

a91727bd

Add tensordot to ATen XLA tensor

asuhan committed 6 years ago

8264616b

Newer Older