7 Commits (c144261d40ab7c5d24e29c03155310a53d79909e)

Author	SHA1	Message	Date
qingqing01	d760b6a58d	Refine the activation type getting in the LSTM operator to speed.	8 years ago
QI JUN	61ec0b9516	Refine device context (#6433) ... There are mainly following fixes: - take `DeviceContext` as the template parameter of math functors and OpKernel instead of `Place` - remove `eigen_device` interface in base class `DeviceContext` - remove `GetEigenDevice` interface in `ExecutionContext` and base class `DeviceContext` - remove unused `platform::EigenDeviceConverter` - rename `REGISTER_OP_GPU_KERNEL` to `REGISTER_OP_CUDA_KERNEL` - rename `USE_GPU_ONLY_OP` to `USE_CUDA_ONLY_OP`	8 years ago
qingqing01	e5b51c4d10	Make lstm_op follow google code style.	8 years ago
Yu Yang	65906ef1d0	Several Enhancement	8 years ago
dangqingqing	d1fbf50b9e	Add unit testing for forwad implementation.	8 years ago
dangqingqing	2a8dbd130d	LSTM Operator forward implementation.	8 years ago
dangqingqing	3cace73701	Add lstm implementation.	8 years ago