Closed TheShy-Dream closed 1 year ago
You're supposed to specify your own input dimension when you instantiate the network, so the 6999192 is not relevant. Also the x.ndim == 4 is for when we were testing with 4-dimensional inputs, which is not the use-case we actually used for MindEye (which defaults to flattening voxels)
收到,辛苦了,祝万事如意,笑口常开,更上一层楼。
"lin0" layer has input and output dimensions that do not match. The input is 699192, but the first layer of "lin0" has a dimension of 15724. I want to know why.