pytorch/xla

Pull Requests Commits

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

Dump graphs to file as in multi-threaded mode outputs get intermingled.

dlibenzi committed 6 years ago

e5db8922

Merge pull request #540 from pytorch/unpin_mkl

dlibenzi committed 6 years ago

Verified 28e02875

Merge pull request #541 from pytorch/debug_graph1

dlibenzi committed 6 years ago

Verified 4ca3cfe8

Remove the pin on MKL as PT updated its code.

dlibenzi committed 6 years ago

b0d61c84

Added graph debug information.

dlibenzi committed 6 years ago

74277a5e

Merge pull request #539 from pytorch/bench_refa1

dlibenzi committed 6 years ago

Verified 2149b7c7

Restructure bench framework to use classes and allow warmup runs.

dlibenzi committed 6 years ago

4aa4eaf5

Merge pull request #538 from pytorch/bench_bw

dlibenzi committed 6 years ago

Verified 01f1e46d

Merge pull request #537 from pytorch/promotem_all

dlibenzi committed 6 years ago

Verified 7fe352b8

Add benchmark with transfer in loop.

dlibenzi committed 6 years ago

524829c5

Do shape/type promotion for binary ops.

dlibenzi committed 6 years ago

d694e313

More concise error message in CheckIntList

asuhan committed 6 years ago

88736aa2

Clean up lambda captures in pooling operations

asuhan committed 6 years ago

93956a9d

Resize kernel, stride and padding for pooling operations

asuhan committed 6 years ago

5c498c90

Support for n-dimensional max pooling

asuhan committed 6 years ago

13bbaa06

Support for n-dimensional average pooling

asuhan committed 6 years ago

862aceb7

Merge pull request #532 from pytorch/add_py_bench

dlibenzi committed 6 years ago

Verified 2509d5e7

Added Python benchmark to measure Cloud TPU latency testing.

dlibenzi committed 6 years ago

02a01396

Older