Newer
Older
function train!(diff, loss, model, W, X, y, η, num_iters, losses = [loss(model(W, X), y)])
for _ in 1:num_iters
∇ = diff(loss, model, W, X, y)
for i in eachindex(W)
W[i] .= W[i] .- η .* ∇[i]
end
push!(losses, loss(model(W, X), y))
end
return losses
end