• CPU / GPU interoperability
• Conversions
• Conversions in Python
• Passing in Pytorch tensors
• The GpuResources object
• Scratch memory
• CUDA streams
• Implemented indexes
• Limitations
• in Python
• Using multiple GPUs
• Performance
• Supported GPUs