Merge pull request #86 from JuliaGNI/update-dependencies

Update dependencies
JuliaGNI · Nov 7, 2023 · 9445c71 · 9445c71
2 parents 6c41e71 + d04b09a
commit 9445c71
Show file tree

Hide file tree

Showing 8 changed files with 17 additions and 25 deletions.
diff --git a/Project.toml b/Project.toml
@@ -36,24 +36,24 @@ Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
 
 [compat]
 AbstractNeuralNetworks = "0.1"
-BandedMatrices = "0.17"
-CUDA = "4"
+BandedMatrices = "0.17, 1"
+CUDA = "4, 5"
 ChainRules = "1"
 ChainRulesCore = "1"
 ChainRulesTestUtils = "1"
 Distances = "0.10"
-Documenter = "0.27"
+Documenter = "0.27, 1"
 ForwardDiff = "0.10"
-GPUArrays = "8"
-GeometricBase = "0.8"
-GeometricEquations = "0.12"
-GeometricIntegrators = "0.12"
-HDF5 = "0.16"
+GPUArrays = "8, 9"
+GeometricBase = "0.9"
+GeometricEquations = "0.14"
+GeometricIntegrators = "0.13"
+HDF5 = "0.16, 0.17"
 KernelAbstractions = "0.9"
 Lux = "0.4, 0.5"
 NLsolve = "4"
 NNlib = "0.8, 0.9"
-Optimisers = "0.2"
+Optimisers = "0.2, 0.3"
 ProgressMeter = "1"
 SafeTestsets = "0.1"
 StatsBase = "0.33, 0.34"

diff --git a/src/GeometricMachineLearning.jl b/src/GeometricMachineLearning.jl
@@ -33,7 +33,7 @@ module GeometricMachineLearning
     import AbstractNeuralNetworks: params, architecture, model, dim
     # export params, architetcure, model
     export dim
-    import GeometricIntegrators.Integrators: method
+    import GeometricIntegrators.Integrators: method, GeometricIntegrator
     import NNlib: σ, sigmoid, softmax
     #import LogExpFunctions: softmax
 

diff --git a/src/data_loader/data_loader.jl b/src/data_loader/data_loader.jl
@@ -18,7 +18,7 @@ For drawing the batch, the sampling is done over n_params and n_time_steps (here
 
 If for the output we have a tensor whose second axis has length 1, we still store it as a tensor and not a matrix. This is because it is not necessarily of length 1. 
 
-TODO: Implement DataLoader that works well with GeometricEnsembles etc.
+TODO: Implement DataLoader that works well with EnsembleProblems etc.
 """
 struct DataLoader{T, AT<:AbstractArray{T}, OT<:Union{AbstractArray, Nothing}}
     input::AT

diff --git a/src/integrator/problem_hnn.jl b/src/integrator/problem_hnn.jl
@@ -20,7 +20,7 @@ function HNNProblem(nn::NeuralNetwork{<:HamiltonianNeuralNetwork}, tspan, tstep,
     HODEProblem(v, f, hamiltonian, tspan, tstep, ics; kwargs...)
 end
 
-function HNNProblem(nn::NeuralNetwork{<:HamiltonianNeuralNetwork}, tspan, tstep, q₀::State, p₀::State; kwargs...)
+function HNNProblem(nn::NeuralNetwork{<:HamiltonianNeuralNetwork}, tspan, tstep, q₀::StateVariable, p₀::StateVariable; kwargs...)
     ics = (q = q₀, p = p₀)
     HNNProblem(nn, tspan, tstep, ics; kwargs...)
 end

diff --git a/src/integrator/problem_lnn.jl b/src/integrator/problem_lnn.jl
@@ -31,11 +31,11 @@ function LNNProblem(nn::NeuralNetwork{<:LagrangianNeuralNetwork}, tspan::Tuple,
     LNNProblem(nn, GeometricEquations._lode_default_g, tspan, tstep, ics; kwargs...)
 end
 
-function LNNProblem(nn::NeuralNetwork{<:LagrangianNeuralNetwork}, g, tspan::Tuple, tstep::Real, q₀::State, p₀::State, λ₀::State = zero(q₀); kwargs...)
+function LNNProblem(nn::NeuralNetwork{<:LagrangianNeuralNetwork}, g, tspan::Tuple, tstep::Real, q₀::StateVariable, p₀::StateVariable, λ₀::StateVariable = zero(q₀); kwargs...)
     ics = (q = q₀, p = p₀, λ = λ₀)
     LNNProblem(nn, g, tspan, tstep, ics; kwargs...)
 end
 
-function LNNProblem(nn::NeuralNetwork{<:LagrangianNeuralNetwork}, tspan::Tuple, tstep::Real, q₀::State, p₀::State, λ₀::State = zero(q₀); kwargs...)
+function LNNProblem(nn::NeuralNetwork{<:LagrangianNeuralNetwork}, tspan::Tuple, tstep::Real, q₀::StateVariable, p₀::StateVariable, λ₀::StateVariable = zero(q₀); kwargs...)
     LNNProblem(nn, GeometricEquations._lode_default_g, tspan, tstep, q₀, p₀, λ₀; kwargs...)
 end
diff --git a/src/integrator/sympnet_integrator.jl b/src/integrator/sympnet_integrator.jl
@@ -10,15 +10,12 @@ function method(nns::NeuralNetSolution{<: NeuralNetwork{<:SympNet}})
 end
 
 
-const IntegratorSympNet{DT,TT} = Integrator{<:Union{HODEProblem{DT,TT}}, <:SympNetMethod}
-
-
-function GeometricIntegrators.integrate(nns::NeuralNetSolution; kwargs...)
+function GeometricIntegrators.Integrators.integrate(nns::NeuralNetSolution; kwargs...)
     integrate(problem(nns), method(nns); kwargs...)
 end
 
 
-function GeometricIntegrators.integrate_step!(int::IntegratorSympNet)
+function GeometricIntegrators.Integrators.integrate_step!(int::GeometricIntegrator{<:SympNetMethod, <:AbstractProblemPODE})
 
     # compute how may times to compose nn ()
     @assert  GeometricIntegrators.Integrators.method(int).Δt % GeometricIntegrators.timestep(int) == 0 

diff --git a/src/nnsolution/neural_net_solution.jl b/src/nnsolution/neural_net_solution.jl
@@ -29,7 +29,7 @@ update_history(nns::NeuralNetSolution, sg::SingleHistory) = _add(nns.history, sg
 
 @inline nn(nns::NeuralNetSolution) = nns.nn
 @inline problem(nns::NeuralNetSolution) = nns.problem
-@inline problem(nns::NeuralNetSolution{T,<:GeometricEnsemble} where T) = GeometricEquations.problem(nns.problem,1)
+@inline problem(nns::NeuralNetSolution{T,<:EnsembleProblem} where T) = GeometricEquations.problem(nns.problem,1)
 @inline GeometricBase.tstep(nns::NeuralNetSolution) = nns.tstep
 @inline loss(nns::NeuralNetSolution) = nns.loss
 

diff --git a/test/integrator/test_integrator.jl b/test/integrator/test_integrator.jl
@@ -63,8 +63,3 @@ training_parameters = TrainingParameters(nruns, method, mopt; batch_size = batch
 neural_net_solution = train!(neuralnet, training_data, training_parameters)
 
 @testnoerror sol = integrate(neural_net_solution)
-
-
-
-
-
Original file line number	Diff line number	Diff line change
Expand Up		@@ -63,8 +63,3 @@ training_parameters = TrainingParameters(nruns, method, mopt; batch_size = batch
		neural_net_solution = train!(neuralnet, training_data, training_parameters)

		@testnoerror sol = integrate(neural_net_solution)