Merge pull request #649 from SciML/Vaibhavdixit02-patch-4

Pass state to OptimizationOptimisers callback

lib/OptimizationOptimisers/src/OptimizationOptimisers.jl

@@ -63,18 +63,18 @@ function SciMLBase.__solve(cache::OptimizationCache{
     Optimization.@withprogress cache.progress name="Training" begin
         for (i, d) in enumerate(data)
             cache.f.grad(G, θ, d...)
-            x = cache.f(θ, cache.p, d...)
+            x = (cache.f(θ, cache.p, d...), state, i)
             cb_call = cache.callback(θ, x...)
             if !(cb_call isa Bool)
-                error("The callback should return a boolean `halt` for whether to stop the optimization process. Please see the sciml_train documentation for information.")
+                error("The callback should return a boolean `halt` for whether to stop the optimization process. Please see the `solve` documentation for information.")
             elseif cb_call
                 break
             end
-            msg = @sprintf("loss: %.3g", x[1])
+            msg = @sprintf("loss: %.3g", first(x)[1])
             cache.progress && ProgressLogging.@logprogress msg i/maxiters
 
             if cache.solver_args.save_best
-                if first(x) < first(min_err)  #found a better solution
+                if first(x)[1] < first(min_err)[1]  #found a better solution
                     min_opt = opt
                     min_err = x
                     min_θ = copy(θ)
@@ -93,7 +93,7 @@ function SciMLBase.__solve(cache::OptimizationCache{
 
     t1 = time()
 
-    SciMLBase.build_solution(cache, cache.opt, θ, x[1], solve_time = t1 - t0)
+    SciMLBase.build_solution(cache, cache.opt, θ, first(x)[1], solve_time = t1 - t0)
     # here should be build_solution to create the output message
 end
 

lib/OptimizationOptimisers/test/runtests.jl

@@ -50,4 +50,20 @@ using Zygote
         sol = Optimization.solve!(cache)
         @test sol.u≈[2.0] atol=1e-3
     end
+
+    @testset "callback" begin
+        rosenbrock(x, p) = (p[1] - x[1])^2 + p[2] * (x[2] - x[1]^2)^2
+        x0 = zeros(2)
+        _p = [1.0, 100.0]
+        l1 = rosenbrock(x0, _p)
+
+        optprob = OptimizationFunction(rosenbrock, Optimization.AutoZygote())
+
+        prob = OptimizationProblem(optprob, x0, _p)
+        function callback(θ, l, state, iter)
+            Optimisers.adjust!(state, 0.1/iter)
+            return false
+        end
+        sol = solve(prob, Optimisers.Adam(0.1), maxiters = 1000, progress = false, callback = callback)
+    end
 end

test/diffeqfluxtests.jl

@@ -87,7 +87,7 @@ function loss_neuralode(p)
 end
 
 iter = 0
-callback = function (p, l, pred)
+callback = function (p, l, pred, args...)
     global iter
     iter += 1
 

test/minibatch.jl

@@ -18,7 +18,7 @@ function dudt_(u, p, t)
     ann(u, p, st)[1] .* u
 end
 
-callback = function (p, l, pred; doplot = false) #callback function to observe training
+callback = function (p, l, pred, args...; doplot = false) #callback function to observe training
     display(l)
     # plot current prediction against data
     if doplot