Conv2d type, and using L1 unstructured pruning to clip 50% (0.5) of the weight tensor (nn.Conv2d layers have two tensors, a weight and a bias) ...
確定! 回上一頁