Whenever I load the onnx file for the first time/run the get_one_face().get(frame) function for the first time on a GPU/CUDA environment, it takes 100+ seconds to load. For further references however, it doesn't take any time at all. The most perplexing part of this is that if I try to load it on a CPU, it will be almost instantaneous. Why is this happening and how do I fix this?
Whenever I load the onnx file for the first time/run the get_one_face().get(frame) function for the first time on a GPU/CUDA environment, it takes 100+ seconds to load. For further references however, it doesn't take any time at all. The most perplexing part of this is that if I try to load it on a CPU, it will be almost instantaneous. Why is this happening and how do I fix this?