Python and C++ source code for moving tensors from CPU to GPU

1

I’m trying to find the source code of the operation to move tensors from CPU to GPU in PyTorch:

data = data.to(device)

However, I’ve checked for many places on Github, but cannot find both the Python code and C++ code for this operation. Does anyone know where the function calls and interfaces for moving tensors between devices are defined? Any help would be appreciated!

1 Like
2

You can find the implementation of to() here, which should eventually call this for a device to device copy and this for a host to device or vice versa copy.

1 Like