Describe the requested improvement
Provide a strategy to ensuring best memory usage for torch models in CUDA.
Associated sits API functionsits_classify()
Additional context
Torch allocates a fixed amount of memory in CUDA which depends on model size. For this reason, SITS needs to find out the size of each model to allocate maximum amount of workers in CUDA.
Describe the requested improvement Provide a strategy to ensuring best memory usage for torch models in CUDA.
Associated sits API function
sits_classify()
Additional context Torch allocates a fixed amount of memory in CUDA which depends on model size. For this reason, SITS needs to find out the size of each model to allocate maximum amount of workers in CUDA.