Merge pull request #652 from SciML/Vaibhavdixit02-patch-4

[WIP]Stats and State

Author: Vaibhavdixit02

Project.toml

@@ -1,6 +1,6 @@
 name = "Optimization"
 uuid = "7f7a1694-90dd-40f0-9382-eb1efda571ba"
-version = "3.20.2"
+version = "3.21.0"
 
 [deps]
 ADTypes = "47edcb42-4c32-4615-8424-f2b9edc5f35b"
@@ -57,7 +57,7 @@ Printf = "1.9"
 ProgressLogging = "0.1"
 Reexport = "1.2"
 ReverseDiff = "1.14"
-SciMLBase = "2.11"
+SciMLBase = "2.16.3"
 SparseArrays = "1.9, 1.10"
 SparseDiffTools = "2.14"
 SymbolicIndexingInterface = "0.3"

lib/OptimizationBBO/Project.toml

@@ -1,7 +1,7 @@
 name = "OptimizationBBO"
 uuid = "3e6eede4-6085-4f62-9a71-46d9bc1eb92b"
 authors = ["Vaibhav Dixit <[email protected]> and contributors"]
-version = "0.1.5"
+version = "0.2.0"
 
 [deps]
 BlackBoxOptim = "a134a8b2-14d6-55f6-9291-3336d3ab0209"
@@ -10,7 +10,7 @@ Reexport = "189a3867-3050-52da-a836-e630ba90ab69"
 
 [compat]
 BlackBoxOptim = "0.6"
-Optimization = "3.15"
+Optimization = "3.21"
 Reexport = "1.2"
 julia = "1"
 

lib/OptimizationBBO/src/OptimizationBBO.jl

@@ -115,7 +115,10 @@ function SciMLBase.__solve(cache::Optimization.OptimizationCache{
         if cache.callback === Optimization.DEFAULT_CALLBACK
             cb_call = false
         else
-            cb_call = cache.callback(decompose_trace(trace, cache.progress), x...)
+            n_steps = BlackBoxOptim.num_steps(trace)
+            curr_u = decompose_trace(trace, cache.progress)
+            opt_state = Optimization.OptimizationState(iteration = n_steps, u = curr_u, objective = x[1], solver_state = trace)
+            cb_call = cache.callback(opt_state, x...)
         end
 
         if !(cb_call isa Bool)
@@ -175,11 +178,13 @@ function SciMLBase.__solve(cache::Optimization.OptimizationCache{
     t1 = time()
 
     opt_ret = Symbol(opt_res.stop_reason)
-
+    stats = Optimization.OptimizationStats(; iterations = opt_res.iterations, time = t1 - t0, fevals = opt_res.f_calls)
     SciMLBase.build_solution(cache, cache.opt,
         BlackBoxOptim.best_candidate(opt_res),
-        BlackBoxOptim.best_fitness(opt_res); original = opt_res,
-        retcode = opt_ret, solve_time = t1 - t0)
+        BlackBoxOptim.best_fitness(opt_res);
+        original = opt_res,
+        retcode = opt_ret,
+        stats = stats)
 end
 
 end

lib/OptimizationBBO/test/runtests.jl

@@ -25,8 +25,8 @@ using Test
     @test 10 * sol.objective < l1
 
     fitness_progress_history = []
-    function cb(best_candidate, fitness)
-        push!(fitness_progress_history, [best_candidate, fitness])
+    function cb(state, fitness)
+        push!(fitness_progress_history, [state.u, fitness])
         return false
     end
     sol = solve(prob, BBO_adaptive_de_rand_1_bin_radiuslimited(), callback = cb)

lib/OptimizationCMAEvolutionStrategy/Project.toml

@@ -1,7 +1,7 @@
 name = "OptimizationCMAEvolutionStrategy"
 uuid = "bd407f91-200f-4536-9381-e4ba712f53f8"
 authors = ["Vaibhav Dixit <[email protected]> and contributors"]
-version = "0.1.4"
+version = "0.2.0"
 
 [deps]
 CMAEvolutionStrategy = "8d3b24bd-414e-49e0-94fb-163cc3a3e411"
@@ -11,7 +11,7 @@ Reexport = "189a3867-3050-52da-a836-e630ba90ab69"
 [compat]
 julia = "1"
 CMAEvolutionStrategy = "0.2"
-Optimization = "3.15"
+Optimization = "3.21"
 Reexport = "1.2"
 
 [extras]

lib/OptimizationCMAEvolutionStrategy/src/OptimizationCMAEvolutionStrategy.jl

@@ -9,7 +9,6 @@ export CMAEvolutionStrategyOpt
 struct CMAEvolutionStrategyOpt end
 
 SciMLBase.allowsbounds(::CMAEvolutionStrategyOpt) = true
-SciMLBase.allowscallback(::CMAEvolutionStrategyOpt) = false #looks like `logger` kwarg can be used to pass it, so should be implemented
 SciMLBase.supports_opt_cache_interface(opt::CMAEvolutionStrategyOpt) = true
 
 function __map_optimizer_args(prob::OptimizationCache, opt::CMAEvolutionStrategyOpt;
@@ -23,7 +22,7 @@ function __map_optimizer_args(prob::OptimizationCache, opt::CMAEvolutionStrategy
     end
 
     mapped_args = (; lower = prob.lb,
-        upper = prob.ub)
+        upper = prob.ub, logger = CMAEvolutionStrategy.BasicLogger(prob.u0; verbosity = 0, callback = callback))
 
     if !isnothing(maxiters)
         mapped_args = (; mapped_args..., maxiter = maxiters)
@@ -74,12 +73,18 @@ function SciMLBase.__solve(cache::OptimizationCache{
 
     cur, state = iterate(cache.data)
 
-    function _cb(trace)
-        cb_call = cache.callback(decompose_trace(trace).metadata["x"], trace.value...)
+    function _cb(opt, y, fvals, perm)
+        curr_u = opt.logger.xbest[end]
+        opt_state = Optimization.OptimizationState(; iteration = length(opt.logger.fmedian),
+            u = curr_u,
+            objective = opt.logger.fbest[end],
+            solver_state = opt.logger)
+
+        cb_call = cache.callback(opt_state, x...)
         if !(cb_call isa Bool)
             error("The callback should return a boolean `halt` for whether to stop the optimization process.")
         end
-        cur, state = iterate(data, state)
+        cur, state = iterate(cache.data, state)
         cb_call
     end
 
@@ -100,11 +105,12 @@ function SciMLBase.__solve(cache::OptimizationCache{
     t1 = time()
 
     opt_ret = opt_res.stop.reason
-
+    stats = Optimization.OptimizationStats(; iterations = length(opt_res.logger.fmedian), time = t1 - t0, fevals = length(opt_res.logger.fmedian))
     SciMLBase.build_solution(cache, cache.opt,
         opt_res.logger.xbest[end],
         opt_res.logger.fbest[end]; original = opt_res,
-        retcode = opt_ret, solve_time = t1 - t0)
+        retcode = opt_ret,
+        stats = stats)
 end
 
 end

lib/OptimizationCMAEvolutionStrategy/test/runtests.jl

@@ -10,4 +10,12 @@ using Test
     prob = OptimizationProblem(f, x0, _p, lb = [-1.0, -1.0], ub = [0.8, 0.8])
     sol = solve(prob, CMAEvolutionStrategyOpt())
     @test 10 * sol.objective < l1
+
+    function cb(state, args...)
+        if state.iteration %10 == 0
+            println(state.u)
+        end
+        return false
+    end
+    sol = solve(prob, CMAEvolutionStrategyOpt(), callback = cb, maxiters = 100)
 end

lib/OptimizationEvolutionary/Project.toml

@@ -1,7 +1,7 @@
 name = "OptimizationEvolutionary"
 uuid = "cb963754-43f6-435e-8d4b-99009ff27753"
 authors = ["Vaibhav Dixit <[email protected]> and contributors"]
-version = "0.1.3"
+version = "0.2.0"
 
 [deps]
 Evolutionary = "86b6b26d-c046-49b6-aa0b-5f0f74682bd6"
@@ -11,7 +11,7 @@ Reexport = "189a3867-3050-52da-a836-e630ba90ab69"
 [compat]
 julia = "1"
 Evolutionary = "0.11"
-Optimization = "3.15"
+Optimization = "3.21"
 Reexport = "1.2"
 
 [extras]

lib/OptimizationEvolutionary/src/OptimizationEvolutionary.jl

@@ -84,7 +84,12 @@ function SciMLBase.__solve(cache::OptimizationCache{
     cur, state = iterate(cache.data)
 
     function _cb(trace)
-        cb_call = cache.callback(decompose_trace(trace).metadata["x"], trace.value...)
+        curr_u = decompose_trace(trace).metadata["x"][end]
+        opt_state = Optimization.OptimizationState(; iteration = decompose_trace(trace).iteration,
+            u = curr_u,
+            objective = x[1],
+            solver_state = trace)
+        cb_call = cache.callback(opt_state, trace.value...)
         if !(cb_call isa Bool)
             error("The callback should return a boolean `halt` for whether to stop the optimization process.")
         end
@@ -127,11 +132,13 @@ function SciMLBase.__solve(cache::OptimizationCache{
     end
     t1 = time()
     opt_ret = Symbol(Evolutionary.converged(opt_res))
-
+    stats = Optimization.OptimizationStats(; iterations = opt_res.iterations
+        , time = t1 - t0, fevals = opt_res.f_calls)
     SciMLBase.build_solution(cache, cache.opt,
         Evolutionary.minimizer(opt_res),
         Evolutionary.minimum(opt_res); original = opt_res,
-        retcode = opt_ret, solve_time = t1 - t0)
+        retcode = opt_ret,
+        stats = stats)
 end
 
 end

lib/OptimizationEvolutionary/test/runtests.jl

@@ -34,4 +34,12 @@ Random.seed!(1234)
     res = zeros(1)
     cons_circ(res, sol.u, nothing)
     @test sol.objective < l1
+
+    function cb(state, args...)
+        if state.iteration %10 == 0
+            println(state.u)
+        end
+        return false
+    end
+    sol = solve(prob, CMAES(μ = 40, λ = 100), callback = cb, maxiters = 100)
 end