fix NNConv docs (#488)

* fix nnconv docstring

* cleanup

src/layers/conv.jl

@@ -659,8 +659,8 @@ For convenience, also functions returning a single `(out*in, num_edges)` matrix
 
 # Arguments
 
-- `in`: The dimension of input features.
-- `out`: The dimension of output features.
+- `in`: The dimension of input node features.
+- `out`: The dimension of output node features.
 - `f`: A (possibly learnable) function acting on edge features.
 - `aggr`: Aggregation operator for the incoming messages (e.g. `+`, `*`, `max`, `min`, and `mean`).
 - `σ`: Activation function.
@@ -670,22 +670,26 @@ For convenience, also functions returning a single `(out*in, num_edges)` matrix
 # Examples:
 
 ```julia
+n_in = 3
+n_in_edge = 10
+n_out = 5
+
 # create data
 s = [1,1,2,3]
 t = [2,3,1,1]
-in_channel = 3
-out_channel = 5
-edim = 10
 g = GNNGraph(s, t)
 
 # create dense layer
-nn = Dense(edim => out_channel * in_channel)
+nn = Dense(n_in_edge => n_out * n_in)
 
 # create layer
-l = NNConv(in_channel => out_channel, nn, tanh, bias = true, aggr = +)
+l = NNConv(n_in => n_out, nn, tanh, bias = true, aggr = +)
+
+x = randn(Float32, n_in, g.num_nodes)
+e = randn(Float32, n_in_edge, g.num_edges)
 
 # forward pass
-y = l(g, x)   
+y = l(g, x, e)  
 ```
 """
 struct NNConv{W, B, NN, F, A} <: GNNLayer