Fix deprecation warnings

Author: claude

src/DeepBSDE.jl

@@ -101,7 +101,6 @@ function DiffEqBase.solve(
     p = prob.p isa AbstractArray ? prob.p : Float32[]
     A = haskey(kwargs, :A) ? prob.A : nothing
     u_domain = prob.x0_sample
-    data = Iterators.repeated((), maxiters)
 
     #hidden layer
     opt = pdealg.opt
@@ -182,15 +181,18 @@ function DiffEqBase.solve(
 
     iters = eltype(x0)[]
     losses = eltype(x0)[]
-    cb = function ()
+    verbose && println("DeepBSDE")
+    for _ in 1:maxiters
+        gs = Flux.gradient(ps) do
+            loss_n_sde()
+        end
+        Flux.Optimise.update!(opt, ps, gs)
         save_everystep && push!(iters, u0(x0)[1])
         l = loss_n_sde()
         push!(losses, l)
         verbose && println("Current loss is: $l")
-        return l < pabstol && Flux.stop()
+        l < pabstol && break
     end
-    verbose && println("DeepBSDE")
-    Flux.train!(loss_n_sde, ps, data, opt; cb = cb)
 
     if !limits
         # Returning iters or simply u0(x0) and the trained neural network approximation u0
@@ -264,13 +266,16 @@ function DiffEqBase.solve(
         loss_() = sum(sol_high()) / trajectories_upper
 
         ps = Flux.params(u0, σᵀ∇u...)
-        cb = function ()
+        opt_upper = Flux.Optimise.Adam(0.01)
+        for _ in 1:maxiters_limits
+            gs = Flux.gradient(ps) do
+                loss_()
+            end
+            Flux.Optimise.update!(opt_upper, ps, gs)
             l = loss_()
-            true && println("Current loss is: $l")
-            return l < 1.0e-6 && Flux.stop()
+            println("Current loss is: $l")
+            l < 1.0e-6 && break
         end
-        dataS = Iterators.repeated((), maxiters_upper)
-        Flux.train!(loss_, ps, dataS, Flux.Optimise.Adam(0.01); cb = cb)
         u_high = loss_()
 
         verbose && println("Lower limit")

src/DeepBSDE_Han.jl

@@ -30,8 +30,6 @@ function DiffEqBase.solve(
     d = length(X0)
     g, f, μ, σ, p = prob.g, prob.f, prob.μ, prob.σ, prob.p
 
-    data = Iterators.repeated((), maxiters)
-
     #hidden layer
     opt = alg.opt
     u0 = alg.u0
@@ -60,16 +58,18 @@ function DiffEqBase.solve(
     iters = eltype(X0)[]
     losses = eltype(X0)[]
 
-    callback = function ()
+    for _ in 1:maxiters
+        gs = Flux.gradient(ps) do
+            loss()
+        end
+        Flux.Optimise.update!(opt, ps, gs)
         save_everystep && push!(iters, u0(X0)[1])
         l = loss()
         push!(losses, l)
         verbose && println("Current loss is: $l")
-        return l < abstol && Flux.stop()
+        l < abstol && break
     end
 
-    Flux.train!(loss, ps, data, opt; cb = callback)
-
     if limits == false
         if save_everystep
             sol = PIDESolution(X0, ts, losses, iters, u0)
@@ -110,13 +110,16 @@ function DiffEqBase.solve(
         loss_() = sum(sol_high()) / trajectories_upper
 
         ps = Flux.params(u0, σᵀ∇u...)
-        callback = function ()
+        opt_limits = Flux.Optimise.Adam(0.01)
+        for _ in 1:maxiters_limits
+            gs = Flux.gradient(ps) do
+                loss_()
+            end
+            Flux.Optimise.update!(opt_limits, ps, gs)
             l = loss_()
             verbose && println("Current loss is: $l")
-            return l < abstol && Flux.stop()
+            l < abstol && break
         end
-        dataS = Iterators.repeated((), maxiters_limits)
-        Flux.train!(loss_, ps, dataS, ADAM(0.01); cb = callback)
         u_high = loss_()
 
         verbose && println("Lower limit")

src/DeepSplitting.jl

@@ -1,7 +1,7 @@
 _copy(t::Tuple) = t
 _copy(t) = t
 function _copy(opt::O) where {O <: Flux.Optimise.AbstractOptimiser}
-    return O([_copy(getfield(opt, f)) for f in fieldnames(typeof(opt))]...)
+    return O([_copy(getfield(opt, fn)) for fn in fieldnames(typeof(opt))]...)
 end
 
 """

src/NNKolmogorov.jl

@@ -13,7 +13,7 @@ struct NNKolmogorov{C, O} <: HighDimPDEAlgorithm
     chain::C
     opt::O
 end
-NNKolmogorov(chain; opt = Flux.ADAM(0.1)) = NNKolmogorov(chain, opt)
+NNKolmogorov(chain; opt = Flux.Adam(0.1)) = NNKolmogorov(chain, opt)
 
 """
 $(TYPEDSIGNATURES)
@@ -69,7 +69,6 @@ function DiffEqBase.solve(
     #hidden layer
     chain = pdealg.chain
     opt = pdealg.opt
-    ps = Flux.params(chain)
     xi = mapreduce(x -> rand(x, 1, trajectories), vcat, xs)
     #Finding Solution to the SDE having initial condition xi. Y = Phi(S(X , T))
     sdeproblem = SDEProblem(

src/NNParamKolmogorov.jl

@@ -14,7 +14,7 @@ struct NNParamKolmogorov{C, O} <: HighDimPDEAlgorithm
     opt::O
 end
 
-NNParamKolmogorov(chain; opt = Flux.ADAM(0.1)) = NNParamKolmogorov(chain, opt)
+NNParamKolmogorov(chain; opt = Flux.Adam(0.1)) = NNParamKolmogorov(chain, opt)
 
 """
 $(TYPEDSIGNATURES)
@@ -75,7 +75,6 @@ function DiffEqBase.solve(
     dps = merge(p_defaults, dps)
 
     chain = pdealg.chain
-    ps = Flux.params(chain)
     opt = pdealg.opt
 
     xi = mapreduce(x -> rand(x, 1, trajectories), vcat, xs)