pytorch/test/cpp/api
Peter Goldsborough 9ce15173fb Move _cudnn_init_dropout_state to TensorOptions and enable cuDNN dropout in C++ API RNNs (#9012)
Summary:
The goal of this PR was to add support for dropout descriptors in the C++ API's RNN class.
The end result is a 4x-5x speedup for our RNN integration tests since they can now use cuDNN instead of autograd when dropout is set.

To achieve this, I had to move `_cudnn_init_dropout_state` to the `TensorOptions` API.

I also fixed a bug around `RNN::cuda()` not flattening parameters for cuDNN.

ebetica ezyang
Closes https://github.com/pytorch/pytorch/pull/9012

Reviewed By: pjh5

Differential Revision: D8689786

Pulled By: goldsborough

fbshipit-source-id: 44fb191f5a38e41c4ded5417306b5bbc012cd56c
2018-06-29 17:25:23 -07:00
..
any.cpp Convert at::Tensor to torch::Tensor in AnyModule (#8968) 2018-06-28 06:40:48 -07:00
cursor.cpp Set random seed at the start of C++ tests (#8903) 2018-06-27 20:09:46 -07:00
integration.cpp Move _cudnn_init_dropout_state to TensorOptions and enable cuDNN dropout in C++ API RNNs (#9012) 2018-06-29 17:25:23 -07:00
main.cpp Split up detail.h (#7836) 2018-05-30 08:55:34 -07:00
misc.cpp Set random seed at the start of C++ tests (#8903) 2018-06-27 20:09:46 -07:00
module.cpp Move _cudnn_init_dropout_state to TensorOptions and enable cuDNN dropout in C++ API RNNs (#9012) 2018-06-29 17:25:23 -07:00
modules.cpp Move _cudnn_init_dropout_state to TensorOptions and enable cuDNN dropout in C++ API RNNs (#9012) 2018-06-29 17:25:23 -07:00
optim.cpp Set random seed at the start of C++ tests (#8903) 2018-06-27 20:09:46 -07:00
optim_baseline.h Use torch:: instead of at:: (#8911) 2018-06-27 14:42:01 -07:00
optim_baseline.py Use torch:: instead of at:: (#8911) 2018-06-27 14:42:01 -07:00
README.md Update C++ API tests to use Catch2 (#7108) 2018-04-30 21:36:35 -04:00
rnn.cpp Move _cudnn_init_dropout_state to TensorOptions and enable cuDNN dropout in C++ API RNNs (#9012) 2018-06-29 17:25:23 -07:00
sequential.cpp Convert at::Tensor to torch::Tensor in AnyModule (#8968) 2018-06-28 06:40:48 -07:00
serialization.cpp Move _cudnn_init_dropout_state to TensorOptions and enable cuDNN dropout in C++ API RNNs (#9012) 2018-06-29 17:25:23 -07:00
static.cpp [C++ API] Make pImpl easy to use in modules to enable happy reference semantics (#8347) 2018-06-18 19:45:53 -07:00
tensor.cpp [C++ API] Bag of fixes (#8843) 2018-06-25 21:11:49 -07:00
tensor_cuda.cpp Make at::tensor faster (#8709) 2018-06-20 14:46:58 -07:00
tensor_options.cpp Created DefaultTensorOptions in ATen (#8647) 2018-06-24 21:15:09 -07:00
tensor_options_cuda.cpp Created DefaultTensorOptions in ATen (#8647) 2018-06-24 21:15:09 -07:00
util.h [C++ API] Better forward methods (#8739) 2018-06-26 13:23:16 -07:00

C++ API Tests

In this folder live the tests for PyTorch's C++ API (formerly known as autogradpp). They use the Catch2 test framework.

CUDA Tests

The way we handle CUDA tests is by separating them into a separate TEST_CASE (e.g. we have optim and optim_cuda test cases in optim.cpp), and giving them the [cuda] tag. Then, inside main.cpp we detect at runtime whether CUDA is available. If not, we disable these CUDA tests by appending ~[cuda] to the test specifications. The ~ disables the tag.

One annoying aspect is that Catch only allows filtering on test cases and not sections. Ideally, one could have a section like LSTM inside the RNN test case, and give this section a [cuda] tag to only run it when CUDA is available. Instead, we have to create a whole separate RNN_cuda test case and put all these CUDA sections in there.

Integration Tests

Integration tests use the MNIST dataset. You must download it by running the following command from the PyTorch root folder:

$ python tools/download_mnist.py -d test/cpp/api/mnist