MFM
MFM copied to clipboard
The article states that the decoder uses a linear layer, but the pre-training code uses Conv2d, and I'm confused about this.
I‘m so glad if you could help me with it.