Error in hyperelasticity demo: function NNlib.σ must be explicitly imported

[Kevin-Mattheus-Moerman]

I just tried to test the hyperelastic demo but get the below error on this line:

σ(∇u) = (1.0/J(F(∇u)))*F(∇u)⋅S(∇u)⋅(F(∇u))'

ERROR: LoadError: error in method definition: function NNlib.σ must be explicitly imported to be extended
Stacktrace:
 [1] top-level scope
   @ none:0

I am using Julia Version 1.6.3 on Ubuntu 20.04 with VScode.

[fverdugo]

Hi @Kevin-Mattheus-Moerman, thanks for reporting!

It seems that some dependency has introduced function σ and it is conflicting with our definition. But I don't understand why...

Are you explicitly using NNlib ? or just running the tutorial verbatim in a fresh Julia session? (this last detail is important)

[Kevin-Mattheus-Moerman]

@fverdugo when I restart the script runs fine. It seems that it fails if I run a script featuring flux.jl first, see below. It features a symbol σ, is that it then? Is there a way to clear definitions before running a script? (as you can see I am new to Julia so did not know these two scripts could "bite" each other in this way.

using CairoMakie, Flux, Statistics

## Auxiliary functions for generating our data
function generate_real_data(n)
    x1 = rand(1,n) .- 0.5
    x2 = (x1 .* x1)*3 .+ randn(1,n)*0.1
    return vcat(x1,x2)
end

function generate_fake_data(n)
    θ  = 2*π*rand(1,n)
    r  = rand(1,n)/3
    x1 = @. r*cos(θ)
    x2 = @. r*sin(θ)+0.5
    return vcat(x1,x2)
end

function NeuralNetwork()
    return Chain(
            Dense(2, 25,relu),
            Dense(25,1,x->σ.(x))
            )
end

## Creating our data
train_size = 5000
real = generate_real_data(train_size)
fake = generate_fake_data(train_size)

## Visualizing
fig1 = Figure()
ax1 = Axis(fig1[1,1],title = "Some dots")
scatter!(ax1,real[1,1:500],real[2,1:500])
scatter!(ax1,fake[1,1:500],fake[2,1:500])
fig1 

# Organizing the data in batches
X    = hcat(real,fake)
Y    = vcat(ones(train_size),zeros(train_size))
data = Flux.Data.DataLoader((X, Y'), batchsize=100,shuffle=true);

# Defining our model, optimization algorithm and loss function
m    = NeuralNetwork()
opt = Descent(0.05)
loss(x, y) = sum(Flux.Losses.binarycrossentropy(m(x), y))

# Training Method 1
ps = Flux.params(m)
epochs = 20
for i in 1:epochs
    Flux.train!(loss, ps, data, opt)
end
println(mean(m(real)),mean(m(fake)))

# Print model prediction# Visualizing the model predictions
fig2 = Figure()
ax2 = Axis(fig2[1,1],title = "Dodged bars with legend 2")
scatter!(ax2,real[1,1:100],real[2,1:100],zcolor=m(real)')
scatter!(ax2,fake[1,1:100],fake[2,1:100],zcolor=m(fake)',legend=false)
fig2

[Kevin-Mattheus-Moerman]

This is resolved thanks