PyTorch Replacing For-Loop Linears with Grouped Conv1d

Question

So I have this simple model:

class ModelSerial(nn.Module):
    def __init__(self, in_features, hidden, out_features):
        super().__init__()
        block