关于求SRE和CONN函数中的参数A的问题
First of all thank you very much to the author for the source code
In the process of learning the source code of SENet, I found that when referring to evaluate function and using subspace_preserving_error function and connectivity function to calculate SRE and CONN results, If you try labels and num_subspaces in other parameters, C_sparse, C_sparse_normalized, and Aff in parameter A, you'll get an AttributeError: layout not found, annotation in torch.norm function code does not norm MPS or sparse tensors, so how to use subspace_preserving_error function and connectivity function to obtain SRE and CONN indexes?
首先非常感谢作者的代码
在学习SENet的源代码过程中，我发现在参考evaluate函数使用subspace_preserving_error函数和connectivity函数计算SRE和CONN结果时，尝试在其他参数位置使用labels和num_subspaces，但在参数A位置使用C_sparse、C_sparse_normalized和Aff传入，都会报错AttributeError: layout not found，torch.norm函数代码内注释不会对MPS或稀疏张量求范数，那么该如何使用subspace_preserving_error函数和connectivity函数求得SRE和CONN指标？
关于求SRE和CONN函数中的参数A的问题
First of all thank you very much to the author for the source code
In the process of learning the source code of SENet, I found that when referring to evaluate function and using subspace_preserving_error function and connectivity function to calculate SRE and CONN results, If you try labels and num_subspaces in other parameters, C_sparse, C_sparse_normalized, and Aff in parameter A, you'll get an AttributeError: layout not found, annotation in torch.norm function code does not norm MPS or sparse tensors, so how to use subspace_preserving_error function and connectivity function to obtain SRE and CONN indexes?
首先非常感谢作者的代码
在学习SENet的源代码过程中，我发现在参考evaluate函数使用subspace_preserving_error函数和connectivity函数计算SRE和CONN结果时，尝试在其他参数位置使用labels和num_subspaces，但在参数A位置使用C_sparse、C_sparse_normalized和Aff传入，都会报错AttributeError: layout not found，torch.norm函数代码内注释不会对MPS或稀疏张量求范数，那么该如何使用subspace_preserving_error函数和connectivity函数求得SRE和CONN指标？