pytorch/xla

Pull Requests Commits

Route scalar values to device data to achieve graph convergence by have scalar leaf nodes being all device data.

dlibenzi committed 6 years ago

929415ae

asuhan committed 6 years ago

af98447b

Route bitwise binary with shape checks to our implementations

asuhan committed 6 years ago

9c1f79f1

Merge pull request #564 from pytorch/fix_jit_breakage_N_plus_1

dlibenzi committed 6 years ago

Verified 7d817be8

Merge pull request #561 from pytorch/lower_scalar_preds

dlibenzi committed 6 years ago

Verified b67b763d

Fix the next JIT build breakage.

dlibenzi committed 6 years ago

be8e6558

Merge pull request #560 from pytorch/index_op_cast

dlibenzi committed 6 years ago

Verified 0a6dcb30

Merge pull request #557 from pytorch/binary_ops_promoted_shape_fix

dlibenzi committed 6 years ago

Verified 85b73201

Handle xla::PRED as scalar.

dlibenzi committed 6 years ago

fa8d47e3

Cast source values into destination type for index operations.

dlibenzi committed 6 years ago

c484bef2

Fix binary ops shape inference.

dlibenzi committed 6 years ago

32606534

Merge pull request #556 from pytorch/fix_build_breakage_N

dlibenzi committed 6 years ago

Verified 49dee0ab

Fix the next build breakage WRT the JIT APIs.

dlibenzi committed 6 years ago

73606033

Fix build after PyTorch core changes

asuhan committed 6 years ago

f76ac65b

Check shapes in equality / close test functions

asuhan committed 6 years ago

37c6df4e

Fix shape of unary min / max

asuhan committed 6 years ago

50bcf7de

Fix output shape of a permutation view

asuhan committed 6 years ago

fe9941f6

Merge pull request #553 from pytorch/grab_optimizer_state_in_sync

dlibenzi committed 6 years ago

Verified 56c5e8fd

Grab the optimizer's state during sync.

dlibenzi committed 6 years ago

7cc8b15c

Merge pull request #554 from pytorch/asuhan/fix_build

dlibenzi committed 6 years ago

Verified 8dde63f8

Fix build after PyTorch core changes

asuhan committed 6 years ago

f4361d8e

Add thnn_conv_transpose2d_backward to ATen XLA tensor

asuhan committed 6 years ago

43a4384c

Handle undefined weight and bias in batch norm

asuhan committed 6 years ago

69225e70

Take the absolute value of a norm argument

asuhan committed 6 years ago

a6faaf2b

Fix build breakage due to at::cross signature change

asuhan committed 6 years ago

f33145b3

Add _softmax to ATen XLA type

asuhan committed 6 years ago

989f2718

Added aten::dropout_ operator.

dlibenzi committed 6 years ago

9ffa72ee

Fix build after latest PyTorch core changes

asuhan committed 6 years ago

5d127be8

Add 2D transposed convolution to ATen XLA tensor

asuhan committed 6 years ago

fb970d46

Merge pull request #543 from pytorch/graphs_to_file

dlibenzi committed 6 years ago

Verified f7c048a6

Newer Older