TopArray is a Python module designed to facilitate the interaction between Python/PyTorch tensor operations and TouchDesigner TOPs. It provides a workflow for referencing CUDA memory from TOPs as tensors and for copying a tensor's data to a scriptTOP efficiently.
To use TopArray, ensure you have TD compatible Python and PyTorch with CUDA installed. The Python installation/environment needs to be visible to TouchDesigner.
Use git clone https://github.com/IntentDev/TopArray.git
or download zip file of the repository.
Load ExampleTensorIO.toe and toggle the Active parameter on the TensorIO componenent.
See component network and TorchExt.py for details on how to interact with the TopArrayInterface and the TopCUDAInterface.
A wrapper class for a TouchDesigner TOP that implements the __cuda_array_interface__
protocol, allowing direct integration with CUDA-aware libraries.
- top (
td.TOP
): The TOP to wrap. - stream (
int
, optional): The CUDA stream to use for transfers. Defaults to0
.
Updates the __cuda_array_interface__
with memory from the TOP.
- Parameters:
stream
: The CUDA stream for transfers. Defaults to0
.
Updates the stream
and data
keys in the __cuda_array_interface__
.
Handles copying data from an array or tensor to a TouchDesigner TOP.
- width (
int
): width of the array that will be used to set the image - height (
int
): height of the array that will be used to set the image - num_comps (
int
): number of color components/channels in the array - dtype (
np.dtype
): NumPy data type of the array - stream (
int
, optional): The CUDA stream to use for transfers. Defaults to0
.
Copies data from a tensor to the specified ScriptTOP.
- Parameters:
scriptOP
(td.scriptOP
): The ScriptTOP to copy the tensor to.
Wraps the copyCUDAMemory
method to copy the tensor data into the ScriptTOP's memory.