SciML · utkarsh530 · Apr 1, 2025 · Apr 1, 2025 · Apr 1, 2025
diff --git a/Project.toml b/Project.toml
@@ -1,4 +1,4 @@
-name = "PSOGPU"
+name = "ParallelParticleSwarms"
 uuid = "ab63da0c-63b4-40fa-a3b7-d2cba5be6419"
 authors = ["Utkarsh <[email protected]> and contributors"]
 version = "0.1.0"

diff --git a/README.md b/README.md
@@ -1,15 +1,15 @@
-# PSOGPU
+# ParallelParticleSwarms
 
-[![Build Status](https://github.com/utkarsh530/PSOGPU.jl/actions/workflows/CI.yml/badge.svg?branch=main)](https://github.com/utkarsh530/PSOGPU.jl/actions/workflows/CI.yml?query=branch%3Amain)
-[![codecov](https://codecov.io/gh/utkarsh530/PSOGPU.jl/graph/badge.svg?token=H5U5UAIRXX)](https://codecov.io/gh/utkarsh530/PSOGPU.jl)
+[![Build Status](https://github.com/utkarsh530/ParallelParticleSwarms.jl/actions/workflows/CI.yml/badge.svg?branch=main)](https://github.com/utkarsh530/ParallelParticleSwarms.jl/actions/workflows/CI.yml?query=branch%3Amain)
+[![codecov](https://codecov.io/gh/utkarsh530/ParallelParticleSwarms.jl/graph/badge.svg?token=H5U5UAIRXX)](https://codecov.io/gh/utkarsh530/ParallelParticleSwarms.jl)
 
 Accelerating convex/non-convex optimization with GPUs using Particle-Swarm based methods
 
 Supports generic Julia's SciML interface
 
 ```julia
 
-using PSOGPU, StaticArrays, CUDA
+using ParallelParticleSwarms, StaticArrays, CUDA
 
 lb = @SArray [-1.0f0, -1.0f0, -1.0f0]
 ub = @SArray [10.0f0, 10.0f0, 10.0f0]

diff --git a/benchmarks/CPU_vs_GPU/Manifest.toml b/benchmarks/CPU_vs_GPU/Manifest.toml
@@ -1484,9 +1484,9 @@ git-tree-sha1 = "8b3b19351fa24791f94d7ae85faf845ca1362541"
 uuid = "32165bc3-0280-59bc-8c0b-c33b6203efab"
 version = "4.2.7+0"
 
-[[deps.PSOGPU]]
+[[deps.ParallelParticleSwarms]]
 deps = ["Adapt", "DiffEqGPU", "Enzyme", "ForwardDiff", "KernelAbstractions", "MLDatasets", "NonlinearSolve", "Optimization", "QuasiMonteCarlo", "Random", "SciMLBase", "Setfield", "SimpleChains", "SimpleNonlinearSolve", "StaticArrays"]
-path = "/home/utkarsh530/.julia/dev/PSOGPU"
+path = "/home/utkarsh530/.julia/dev/ParallelParticleSwarms"
 uuid = "ab63da0c-63b4-40fa-a3b7-d2cba5be6419"
 version = "1.0.0-DEV"
 

diff --git a/benchmarks/CPU_vs_GPU/Project.toml b/benchmarks/CPU_vs_GPU/Project.toml
@@ -6,5 +6,5 @@ OptimizationBBO = "3e6eede4-6085-4f62-9a71-46d9bc1eb92b"
 OptimizationFlux = "253f991c-a7b2-45f8-8852-8b9a9df78a86"
 OptimizationOptimJL = "36348300-93cb-4f02-beb5-3c3902f8871e"
 OptimizationOptimisers = "42dfb2eb-d2b4-4451-abcd-913932933ac1"
-PSOGPU = "ab63da0c-63b4-40fa-a3b7-d2cba5be6419"
+ParallelParticleSwarms = "ab63da0c-63b4-40fa-a3b7-d2cba5be6419"
 StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
diff --git a/benchmarks/CPU_vs_GPU/benchmark.jl b/benchmarks/CPU_vs_GPU/benchmark.jl
@@ -2,7 +2,7 @@ using Pkg
 
 Pkg.activate(@__DIR__)
 
-using PSOGPU, StaticArrays, KernelAbstractions, Optimization
+using ParallelParticleSwarms, StaticArrays, KernelAbstractions, Optimization
 using CUDA
 
 device!(2)
@@ -51,11 +51,11 @@ sol = solve(prob,
 @show sol.stats.time
 
 sol = solve(prob,
-    PSOGPU.HybridPSO(; backend = CUDABackend(),
-        pso = PSOGPU.ParallelPSOKernel(n_particles;
+    ParallelParticleSwarms.HybridPSO(; backend = CUDABackend(),
+        pso = ParallelParticleSwarms.ParallelPSOKernel(n_particles;
             global_update = false,
             backend = CUDABackend()),
-        local_opt = PSOGPU.LBFGS()), maxiters = 500,
+        local_opt = ParallelParticleSwarms.LBFGS()), maxiters = 500,
     local_maxiters = 30)
 
 @show sol.objective
@@ -77,7 +77,8 @@ for n_particles in Ns
     ## CPU solve
     backend = CPU()
     opt = ParallelSyncPSOKernel(n_particles; backend)
-    init_gbest, particles = PSOGPU.init_particles(prob, opt, typeof(prob.u0))
+    init_gbest, particles = ParallelParticleSwarms.init_particles(
+        prob, opt, typeof(prob.u0))
 
     particles_eltype = eltype(particles) === Float64 ? Float32 : eltype(particles)
 
@@ -87,12 +88,12 @@ for n_particles in Ns
 
     copyto!(backend_particles, particles)
 
-    PSOGPU.vectorized_solve!(prob,
+    ParallelParticleSwarms.vectorized_solve!(prob,
         init_gbest,
         backend_particles,
         opt; maxiters = 500)
 
-    el_time = @elapsed PSOGPU.vectorized_solve!(prob,
+    el_time = @elapsed ParallelParticleSwarms.vectorized_solve!(prob,
         init_gbest,
         backend_particles,
         opt; maxiters = 500)
@@ -112,12 +113,12 @@ for n_particles in Ns
 
     copyto!(backend_particles, particles)
 
-    PSOGPU.vectorized_solve!(prob,
+    ParallelParticleSwarms.vectorized_solve!(prob,
         init_gbest,
         backend_particles,
         opt; maxiters = 500)
 
-    el_time = @elapsed PSOGPU.vectorized_solve!(prob,
+    el_time = @elapsed ParallelParticleSwarms.vectorized_solve!(prob,
         init_gbest,
         backend_particles,
         opt; maxiters = 500)
@@ -129,12 +130,12 @@ for n_particles in Ns
     gpu_init_gbest = KernelAbstractions.allocate(backend, typeof(init_gbest), (1,))
     copyto!(gpu_init_gbest, [init_gbest])
 
-    PSOGPU.vectorized_solve!(prob,
+    ParallelParticleSwarms.vectorized_solve!(prob,
         gpu_init_gbest,
         backend_particles,
         opt, Val(opt.global_update); maxiters = 500)
 
-    el_time = @elapsed PSOGPU.vectorized_solve!(prob,
+    el_time = @elapsed ParallelParticleSwarms.vectorized_solve!(prob,
         gpu_init_gbest,
         backend_particles,
         opt, Val(opt.global_update); maxiters = 500)
@@ -146,12 +147,12 @@ for n_particles in Ns
     gpu_init_gbest = KernelAbstractions.allocate(backend, typeof(init_gbest), (1,))
     copyto!(gpu_init_gbest, [init_gbest])
 
-    PSOGPU.vectorized_solve!(prob,
+    ParallelParticleSwarms.vectorized_solve!(prob,
         gpu_init_gbest,
         backend_particles,
         opt, Val(opt.global_update); maxiters = 500)
 
-    el_time = @elapsed PSOGPU.vectorized_solve!(prob,
+    el_time = @elapsed ParallelParticleSwarms.vectorized_solve!(prob,
         gpu_init_gbest,
         backend_particles,
         opt, Val(opt.global_update); maxiters = 500)
@@ -167,19 +168,19 @@ for n_particles in Ns
     @info n_particles
 
     sol = solve(prob,
-        PSOGPU.HybridPSO(; backend = CUDABackend(),
-            pso = PSOGPU.ParallelPSOKernel(n_particles;
+        ParallelParticleSwarms.HybridPSO(; backend = CUDABackend(),
+            pso = ParallelParticleSwarms.ParallelPSOKernel(n_particles;
                 global_update = false,
                 backend = CUDABackend()),
-            local_opt = PSOGPU.LBFGS()), maxiters = 500,
+            local_opt = ParallelParticleSwarms.LBFGS()), maxiters = 500,
         local_maxiters = 30)
 
     sol = solve(prob,
-        PSOGPU.HybridPSO(; backend = CUDABackend(),
-            pso = PSOGPU.ParallelPSOKernel(n_particles;
+        ParallelParticleSwarms.HybridPSO(; backend = CUDABackend(),
+            pso = ParallelParticleSwarms.ParallelPSOKernel(n_particles;
                 global_update = false,
                 backend = CUDABackend()),
-            local_opt = PSOGPU.LBFGS()), maxiters = 500,
+            local_opt = ParallelParticleSwarms.LBFGS()), maxiters = 500,
         local_maxiters = 30)
 
     push!(gpu_hybrid_times, sol.stats.time)
@@ -309,15 +310,15 @@ using Statistics
 #     push!(gpu_queue_lock_times_total, el_time)
 
 #     sol = solve(prob,
-#         PSOGPU.HybridPSO(; backend = CUDABackend(),
-#         pso = PSOGPU.ParallelPSOKernel(n_particles; global_update = false, backend = CUDABackend()),
-#         local_opt = PSOGPU.LBFGS()), maxiters = 500,
+#         ParallelParticleSwarms.HybridPSO(; backend = CUDABackend(),
+#         pso = ParallelParticleSwarms.ParallelPSOKernel(n_particles; global_update = false, backend = CUDABackend()),
+#         local_opt = ParallelParticleSwarms.LBFGS()), maxiters = 500,
 #         local_maxiters = 30)
 
 #     el_time = @elapsed solve(prob,
-#     PSOGPU.HybridPSO(; backend = CUDABackend(),
-#     pso = PSOGPU.ParallelPSOKernel(n_particles; global_update = false, backend = CUDABackend()),
-#     local_opt = PSOGPU.LBFGS()), maxiters = 500,
+#     ParallelParticleSwarms.HybridPSO(; backend = CUDABackend(),
+#     pso = ParallelParticleSwarms.ParallelPSOKernel(n_particles; global_update = false, backend = CUDABackend()),
+#     local_opt = ParallelParticleSwarms.LBFGS()), maxiters = 500,
 #     local_maxiters = 30)
 
 #     push!(gpu_hybrid_times_total, el_time)

diff --git a/benchmarks/CPU_vs_GPU/wp_algs.jl b/benchmarks/CPU_vs_GPU/wp_algs.jl
@@ -2,7 +2,7 @@ using Pkg
 
 Pkg.activate(@__DIR__)
 
-using PSOGPU, StaticArrays, KernelAbstractions, Optimization
+using ParallelParticleSwarms, StaticArrays, KernelAbstractions, Optimization
 using CUDA
 
 device!(2)
@@ -55,11 +55,11 @@ sol = solve(prob,
 @show sol.stats.time
 
 sol = solve(prob,
-    PSOGPU.HybridPSO(; backend = CUDABackend(),
-        pso = PSOGPU.ParallelPSOKernel(n_particles;
+    ParallelParticleSwarms.HybridPSO(; backend = CUDABackend(),
+        pso = ParallelParticleSwarms.ParallelPSOKernel(n_particles;
             global_update = false,
             backend = CUDABackend()),
-        local_opt = PSOGPU.LBFGS()), maxiters = 500,
+        local_opt = ParallelParticleSwarms.LBFGS()), maxiters = 500,
     local_maxiters = 30)
 
 @show sol.objective
@@ -91,7 +91,7 @@ function solve_run(prob, alg, maxiters; runs = 10, kwargs...)
     # 4 was a good candidate
     Random.seed!(rng, 1)
     for run in 1:runs
-        sol = if alg isa PSOGPU.HybridPSO
+        sol = if alg isa ParallelParticleSwarms.HybridPSO
             solve(prob, alg; maxiters, local_maxiters = 30)
         else
             solve(prob, alg; maxiters, kwargs...)
@@ -135,11 +135,11 @@ for n_particles in Ns
     push!(gpu_queue_lock_times, sol_time)
 
     obj, solve_time = solve_run(prob,
-        PSOGPU.HybridPSO(; backend = CUDABackend(),
-            pso = PSOGPU.ParallelPSOKernel(n_particles;
+        ParallelParticleSwarms.HybridPSO(; backend = CUDABackend(),
+            pso = ParallelParticleSwarms.ParallelPSOKernel(n_particles;
                 global_update = false,
                 backend = CUDABackend()),
-            local_opt = PSOGPU.LBFGS()), 500)
+            local_opt = ParallelParticleSwarms.LBFGS()), 500)
 
     push!(gpu_hybrid_loss, obj)
     push!(gpu_hybrid_times, solve_time)
@@ -256,11 +256,11 @@ sol = solve(prob,
 @show sol.stats.time
 
 sol = solve(prob,
-    PSOGPU.HybridPSO(; backend = CUDABackend(),
-        pso = PSOGPU.ParallelPSOKernel(n_particles;
+    ParallelParticleSwarms.HybridPSO(; backend = CUDABackend(),
+        pso = ParallelParticleSwarms.ParallelPSOKernel(n_particles;
             global_update = false,
             backend = CUDABackend()),
-        local_opt = PSOGPU.LBFGS()), maxiters = 500,
+        local_opt = ParallelParticleSwarms.LBFGS()), maxiters = 500,
     local_maxiters = 30)
 
 @show sol.objective
@@ -300,7 +300,7 @@ function solve_run(prob, alg, maxiters; runs = 10, kwargs...)
     # 4 was a good candidate
     Random.seed!(rng, 9)
     for run in 1:runs
-        sol = if alg isa PSOGPU.HybridPSO
+        sol = if alg isa ParallelParticleSwarms.HybridPSO
             solve(prob, alg; maxiters, local_maxiters = maxiters)
         else
             solve(prob, alg; maxiters, kwargs...)
@@ -377,23 +377,23 @@ begin
         push!(pso_cpu_time, sol_time)
 
         # sol = solve(prob,
-        #     PSOGPU.HybridPSO(; backend = CUDABackend(),
-        #     pso = PSOGPU.ParallelPSOKernel(n_particles; global_update = false, backend = CUDABackend()),
-        #     local_opt = PSOGPU.LBFGS()), maxiters = iters,
+        #     ParallelParticleSwarms.HybridPSO(; backend = CUDABackend(),
+        #     pso = ParallelParticleSwarms.ParallelPSOKernel(n_particles; global_update = false, backend = CUDABackend()),
+        #     local_opt = ParallelParticleSwarms.LBFGS()), maxiters = iters,
         #     local_maxiters = iters)
 
         # sol = solve(prob,
-        #         PSOGPU.HybridPSO(; backend = CUDABackend(),
-        #         pso = PSOGPU.ParallelPSOKernel(n_particles; global_update = false, backend = CUDABackend()),
-        #         local_opt = PSOGPU.LBFGS()), maxiters = iters,
+        #         ParallelParticleSwarms.HybridPSO(; backend = CUDABackend(),
+        #         pso = ParallelParticleSwarms.ParallelPSOKernel(n_particles; global_update = false, backend = CUDABackend()),
+        #         local_opt = ParallelParticleSwarms.LBFGS()), maxiters = iters,
         #         local_maxiters = iters)
 
         obj, solve_time = solve_run(prob,
-            PSOGPU.HybridPSO(; backend = CUDABackend(),
-                pso = PSOGPU.ParallelPSOKernel(n_particles;
+            ParallelParticleSwarms.HybridPSO(; backend = CUDABackend(),
+                pso = ParallelParticleSwarms.ParallelPSOKernel(n_particles;
                     global_update = false,
                     backend = CUDABackend()),
-                local_opt = PSOGPU.LBFGS()), iters)
+                local_opt = ParallelParticleSwarms.LBFGS()), iters)
 
         push!(hybrid_losses, obj)
         push!(hybrid_time, solve_time)

diff --git a/benchmarks/Fitzhugh_Nagumo/Manifest.toml b/benchmarks/Fitzhugh_Nagumo/Manifest.toml
@@ -1515,7 +1515,7 @@ git-tree-sha1 = "8b3b19351fa24791f94d7ae85faf845ca1362541"
 uuid = "32165bc3-0280-59bc-8c0b-c33b6203efab"
 version = "4.2.7+0"
 
-[[deps.PSOGPU]]
+[[deps.ParallelParticleSwarms]]
 deps = ["Adapt", "CUDA", "DiffEqGPU", "Enzyme", "ForwardDiff", "KernelAbstractions", "MLDatasets", "NonlinearSolve", "Optimization", "QuasiMonteCarlo", "Random", "SciMLBase", "Setfield", "SimpleChains", "SimpleNonlinearSolve", "StaticArrays"]
 path = "../.."
 uuid = "ab63da0c-63b4-40fa-a3b7-d2cba5be6419"

diff --git a/benchmarks/Fitzhugh_Nagumo/Project.toml b/benchmarks/Fitzhugh_Nagumo/Project.toml
@@ -4,6 +4,6 @@ BenchmarkTools = "6e4b80f9-dd63-53aa-95a3-0cdb28fa8baf"
 CUDA = "052768ef-5323-5732-b1bb-66c8b64840ba"
 OptimizationOptimJL = "36348300-93cb-4f02-beb5-3c3902f8871e"
 OrdinaryDiffEq = "1dea7af3-3e70-54e6-95c3-0bf5283fa5ed"
-PSOGPU = "ab63da0c-63b4-40fa-a3b7-d2cba5be6419"
+ParallelParticleSwarms = "ab63da0c-63b4-40fa-a3b7-d2cba5be6419"
 SciMLSensitivity = "1ed8b502-d754-442c-8d5d-10ac956f44a1"
 StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
diff --git a/benchmarks/Fitzhugh_Nagumo/ode_estimation.jl b/benchmarks/Fitzhugh_Nagumo/ode_estimation.jl
@@ -1,6 +1,6 @@
 using Pkg
 Pkg.activate(@__DIR__)
-using PSOGPU, OrdinaryDiffEq, StaticArrays
+using ParallelParticleSwarms, OrdinaryDiffEq, StaticArrays
 using SciMLSensitivity, Optimization
 
 function f(u, p, t)
@@ -91,7 +91,7 @@ using OptimizationOptimJL
 
 optprob = OptimizationProblem(loss, prob.p, (prob, t_short); lb = lb, ub = ub)
 
-using PSOGPU
+using ParallelParticleSwarms
 using CUDA
 
 using Random
@@ -101,7 +101,7 @@ rng = Random.default_rng()
 Random.seed!(rng, 0)
 
 opt = ParallelPSOKernel(n_particles)
-gbest, particles = PSOGPU.init_particles(optprob, opt, typeof(prob.u0))
+gbest, particles = ParallelParticleSwarms.init_particles(optprob, opt, typeof(prob.u0))
 
 gpu_data = cu([SVector{length(prob.u0), eltype(prob.u0)}(@view data_short[:, i])
                for i in 1:length(t_short)])
@@ -118,7 +118,7 @@ solver_cache = (; losses, gpu_particles, gpu_data, gbest)
 
 adaptive = false
 
-@time gsol = PSOGPU.parameter_estim_ode!(prob,
+@time gsol = ParallelParticleSwarms.parameter_estim_ode!(prob,
     solver_cache,
     lb,
     ub, Val(adaptive);
@@ -128,7 +128,7 @@ adaptive = false
 
 using BenchmarkTools
 
-@benchmark PSOGPU.parameter_estim_ode!($prob,
+@benchmark ParallelParticleSwarms.parameter_estim_ode!($prob,
     $(deepcopy(solver_cache)),
     $lb,
     $ub, $Val(adaptive);

diff --git a/benchmarks/Lotka_Volterra/Manifest.toml b/benchmarks/Lotka_Volterra/Manifest.toml
@@ -1527,7 +1527,7 @@ git-tree-sha1 = "949347156c25054de2db3b166c52ac4728cbad65"
 uuid = "90014a1f-27ba-587c-ab20-58faa44d9150"
 version = "0.11.31"
 
-[[deps.PSOGPU]]
+[[deps.ParallelParticleSwarms]]
 deps = ["Adapt", "CUDA", "DiffEqGPU", "Enzyme", "ForwardDiff", "KernelAbstractions", "MLDatasets", "NonlinearSolve", "Optimization", "QuasiMonteCarlo", "Random", "SciMLBase", "Setfield", "SimpleChains", "SimpleNonlinearSolve", "StaticArrays"]
 path = "../.."
 uuid = "ab63da0c-63b4-40fa-a3b7-d2cba5be6419"

diff --git a/benchmarks/Lotka_Volterra/Project.toml b/benchmarks/Lotka_Volterra/Project.toml
@@ -5,7 +5,7 @@ CUDA = "052768ef-5323-5732-b1bb-66c8b64840ba"
 OptimizationFlux = "253f991c-a7b2-45f8-8852-8b9a9df78a86"
 OptimizationOptimJL = "36348300-93cb-4f02-beb5-3c3902f8871e"
 OrdinaryDiffEq = "1dea7af3-3e70-54e6-95c3-0bf5283fa5ed"
-PSOGPU = "ab63da0c-63b4-40fa-a3b7-d2cba5be6419"
+ParallelParticleSwarms = "ab63da0c-63b4-40fa-a3b7-d2cba5be6419"
 SciMLBase = "0bca4576-84f4-4d90-8ffe-ffa030f20462"
 SciMLSensitivity = "1ed8b502-d754-442c-8d5d-10ac956f44a1"
 StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"