Torch OpenCLOpenCL Backend for PyTorch

Experimental PyTorch backend using OpenCL and PrivateUse1 integration.

PyTorch Integration

Integrates OpenCL directly into the PyTorch dispatcher using the PrivateUse1 backend system.

Runtime abstraction layer for OpenCL platforms, devices, contexts, and command queues.

Supports tensor allocation, tensor transfer, and tensor execution using OpenCL buffers.

Custom allocator integration between PyTorch storage and OpenCL memory objects.

Executes OpenCL kernels through the PyTorch dispatcher and backend runtime pipeline.

Backend validation is performed using full integration testing against real runtime execution.