conv1d
There is no nn.functional.conv1d in mlx, use mx.core.conv1d.
The weight matrix in pytorch is (out_channels, in_channels, kernel_size),
The weight matrix in mlx is (out_channels, kernel_size, in_channels),
The input in pytorch is (N, C, L).
The input in mlx is (N, L, C)