Merge branch 'dev'

README.md

@@ -45,20 +45,20 @@ In all the examples below, one needs `Flux`, `ChainPlots` and `Plots`, while for
 
 ### Dense and Recurrent layers
 
-```julia
+````julia
 julia> nnr = Chain(Dense(2,5,σ),RNN(5,4,relu), LSTM(4,4), GRU(4,4), Dense(4,3))
 Chain(Dense(2, 5, σ), Recur(RNNCell(5, 4, relu)), Recur(LSTMCell(4, 4)), Dense(4, 3))
 
 julia> plot(nnr, title="With theme default", titlefontsize=10)
-```
+````
 
 ![nnr_default plot](examples/img/nnr_default.png)
 
 ### Variable-input layers
 
 Variable-input functional layers are also accepted. If given as the first layer, then an initial input must be provided, otherwise, the input data is not needed. Here are two examples, illustrating each case.
 
-```julia
+````julia
 julia> dx(x) = x[2:end]-x[1:end-1]
 dx (generic function with 1 method)
 
@@ -69,26 +69,26 @@ julia> nna = Chain(Dense(2,5,σ), dx, RNN(4,6,relu), x³, LSTM(6,4), GRU(4,4), D
 Chain(Dense(2, 5, σ), dx, Recur(RNNCell(4, 6, relu)), x³, Recur(LSTMCell(6, 4)), Recur(GRUCell(4, 4)), Dense(4, 3))
 
 julia> plot(nna, title="$nna", titlefontsize=7)
-```
+````
 
 ![nna plot](examples/img/nna.png)
 
-```julia
+````julia
 julia> nnx = Chain(x³, dx, LSTM(5,10), Dense(10,5))
 Chain(x³, dx, Recur(LSTMCell(5, 10)), Dense(10, 5))
 
 julia> input_data = rand(6);
 
 julia> plot(nnx, input_data, title="$nnx", titlefontsize=9)
-```
+````
 
 ![nnx plot](examples/img/nnx.png)
 
 ### Convolutional networks
 
 A neural network with a one-dimensional convolutional layer:
 
-```julia
+````julia
 julia> reshape6x1x1(a) = reshape(a, 6,  1, 1)
 reshape6x1x1 (generic function with 1 method)
 
@@ -99,27 +99,27 @@ julia> nnrs = Chain(x³, Dense(3,6), reshape6x1x1, Conv((2,), 1=>1), slice, Dens
 Chain(x³, Dense(3, 6), reshape6x1x1, Conv((2,), 1=>1), slice, Dense(5, 4))
 
 julia> plot(nnrs, Float32.(rand(3)), title="$nnrs", titlefontsize=9)
-```
+````
 
 ![nnrs plot](examples/img/nnrs.png)
 
 Now with a two-dimensional convolution:
 
-```julia
+````julia
 julia> reshape4x4x1x1(a) = reshape(a, 4, 4, 1, 1)
 reshape4x4x1x1 (generic function with 1 method)
 
 julia> nnrs2d = Chain(x³, Dense(4,16), reshape4x4x1x1, Conv((2,2), 1=>1), slice)
 Chain(x³, Dense(4, 16), reshape4x4x1x1, Conv((2, 2), 1=>1), slice)
 
 julia> plot(nnrs2d, Float32.(rand(4)), title="$nnrs2d", titlefontsize=9)
-```
+````
 
 ![nnrs2d plot](examples/img/nnrs2d.png)
 
 With convolutional and pooling layers:
 
-```julia
+````julia
 julia> nncp = Chain(
            Conv((3, 3), 1=>2, pad=(1,1), bias=false),
            MaxPool((2,2)),
@@ -138,15 +138,15 @@ Chain(
 )                   # Total: 5 arrays, 242 parameters, 2.047 KiB.
 
 julia> plot(nncp, (16, 16, 1, 1), title="Chain with convolutional and pooling layers", titlefontsize=10)
-```
+````
 
 ![nncp plot](examples/img/nncp.png)
 
 ### From Chain to MetaGraph
 
 With `ChainPlots.chaingraph()` we can convert a `Flux.Chain` to a `MetaGraph`.
 
-```julia
+````julia
 julia> nnr = Chain(Dense(2,5,σ),RNN(5,4,relu), LSTM(4,4), GRU(4,4), Dense(4,3))
 Chain(Dense(2, 5, σ), Recur(RNNCell(5, 4, relu)), Recur(LSTMCell(4, 4)), Recur(GRUCell(4, 4)), Dense(4, 3))
 
@@ -175,13 +175,13 @@ julia> get_prop.(Ref(mg_nnr), 15, [:loc_x, :loc_y])
 2-element Vector{Real}:
  3.0
  0.75
-```
+````
 
 ### Visualizing the MetaGraph
 
 We may visualize the generated MetaGraph with [JuliaGraphs/GraphPlot.jl](https://github.com/JuliaGraphs/GraphPlot.jl). We use the attributes `:loc_x`, `:loc_y`, and `:neuron_color` to properly position and color every neuron.
 
-```julia
+````julia
 julia> nnr = Chain(Dense(2,5,σ),RNN(5,4,relu), LSTM(4,4), GRU(4,4), Dense(4,3))
 Chain(Dense(2, 5, σ), Recur(RNNCell(5, 4, relu)), Recur(LSTMCell(4, 4)), Recur(GRUCell(4, 4)), Dense(4, 3))
 
@@ -222,7 +222,7 @@ julia> nodefillc = [parse(Colorant, get_prop(mg_nnr, v, :neuron_color)) for v in
  RGB{N0f8}(0.565,0.933,0.565)
 
 julia> draw(PNG("img/mg_nnr.png", 600, 400), gplot(mg_nnr, locs_x, locs_y, nodefillc=nodefillc))
-```
+````
 
 And here is the result.