Lucas Lee

Thank you for referencing my project. Could you tell me more in detail what 'inserting some frames between the two' is?

Our code doesn't limit the maximum size. How about checking the computational environment such as GPU memory?

This repo is the pytorch implementation of our research. I did not consider about the onnx export.

CUDNN_STATUS_EXECUTION_FAILED error usually occurs when the CUDA, PyTorch, or graphic card driver versions don't match. Also, CUDA_ERROR_ILLEGAL_ADDRESS error usually occurs when the inputs for the cupy functions are not appropriate...