from this presentation it seems that the NPU supports pruning and is able to take advantage of sparsity. I was curious if the IMX 8M plus supports random sparsity. The ML guide and EIQ tools only mention quantization, they don't talk about pruning so I was curious if random sparsity is also supported in the board or not.
Sparsity is not be fully supported in MX8 since there is no software to do it, you can do pruning and you can see the performance between cpu and npu from the document.
