4 Commits (f1a9efcac5743e05fc4e1dbc53c08f69ec19d5a8)

Author	SHA1	Message	Date
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`	7 years ago
dangqingqing	d432b10d8a	Update cuda kernel and doc.	7 years ago
dangqingqing	5bd1e73f5e	Refine and speedup momentum operator.	7 years ago
sidgoyal78	d28b3094dd	Add momentum operator	7 years ago