PINNErrorVsTime Benchmark Updates #1159
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@ChrisRackauckas I got an error on running the iterations which said that the maxiters are less than 1000 so I set all maxiters to 1100. Actually the decision was a bit arbitrary but is that a good number ?? |
|
https://docs.sciml.ai/SciMLBenchmarksOutput/v0.5/PINNErrorsVsTime/diffusion_et/ It's supposed to just show the error over time and then get cutoff. I don't see why making it longer would help. |
Yes. Actually I set it to that number just to get rid of that error |
|
Wait what's the error? |
the error went off on running again |
|
what error? |
|
maxiters should be a number greater than 1000 |
|
can you please just show the error... |
|
|
I see, that's for the sampling algorithm. You should only need that on Cuhre? |
Yes. But as Cuhre was the first one in the line I thought setting to 1100 just for it would not solve the problem, so I set it to 1100 for all of them |
|
The CI has passed here. And all the code seems to run perfectly. Can you please review ?? |
|
@ArnoStrouwen SciML/Integrals.jl#124 can you remind me what the purpose behind this was? |
|
I don't remember myself, but that PR links to: |
|
Uninitialized memory in the original C: giordano/Cuba.jl#12 (comment) fantastic stuff numerical community, that's your classic method that everyone says when they say "all of the old stuff is robust" 😅 |
|
Can you force latest majors and make sure the manifest resolves? |
|
I bump forced the latest versions and resolved manifests but initially there were a lot of version conflicts. I removed IntegralsCuba and IntegralsCubature for a while to resolve them. The manifest resolved but adding both of them back again poses some more version conflicts |
|
Can you share the resolution errors? |
|
@ChrisRackauckas These are the resolution errors that occur |
|
Oh those were turned into extensions. Change |
|
Sure !! 🫡 |
|
This comes from the phi layer in Neural PDE |
|
What changes did you do? What line is erroring? Please be specific and verbose when you report some issue |
I'm sorry if I wasn't clear and verbose at this point. The issue is the following: using NeuralPDE
using Integrals, Cubature, Cuba
using ModelingToolkit, Optimization, OptimizationOptimJL
using Lux, Plots
using DelimitedFiles
using QuasiMonteCarlo
import ModelingToolkit: Interval, infimum, supremum
function allen_cahn(strategy, minimizer, maxIters)
## DECLARATIONS
@parameters t x1 x2 x3 x4
@variables u(..)
Dt = Differential(t)
Dxx1 = Differential(x1)^2
Dxx2 = Differential(x2)^2
Dxx3 = Differential(x3)^2
Dxx4 = Differential(x4)^2
# Discretization
tmax = 1.0
x1width = 1.0
x2width = 1.0
x3width = 1.0
x4width = 1.0
tMeshNum = 10
x1MeshNum = 10
x2MeshNum = 10
x3MeshNum = 10
x4MeshNum = 10
dt = tmax / tMeshNum
dx1 = x1width / x1MeshNum
dx2 = x2width / x2MeshNum
dx3 = x3width / x3MeshNum
dx4 = x4width / x4MeshNum
domains = [t ∈ Interval(0.0, tmax),
x1 ∈ Interval(0.0, x1width),
x2 ∈ Interval(0.0, x2width),
x3 ∈ Interval(0.0, x3width),
x4 ∈ Interval(0.0, x4width)]
ts = 0.0:dt:tmax
x1s = 0.0:dx1:x1width
x2s = 0.0:dx2:x2width
x3s = 0.0:dx3:x3width
x4s = 0.0:dx4:x4width
# Operators
Δu = Dxx1(u(t, x1, x2, x3, x4)) + Dxx2(u(t, x1, x2, x3, x4)) + Dxx3(u(t, x1, x2, x3, x4)) + Dxx4(u(t, x1, x2, x3, x4)) # Laplacian
# Equation
eq = Dt(u(t, x1, x2, x3, x4)) - Δu - u(t, x1, x2, x3, x4) + u(t, x1, x2, x3, x4) * u(t, x1, x2, x3, x4) * u(t, x1, x2, x3, x4) ~ 0 #ALLEN CAHN EQUATION
initialCondition = 1 / (2 + 0.4 * (x1 * x1 + x2 * x2 + x3 * x3 + x4 * x4)) # see PNAS paper
bcs = [u(0, x1, x2, x3, x4) ~ initialCondition] #from literature
## NEURAL NETWORK
n = 10 #neuron number
chain = Lux.Chain(Lux.Dense(5, n, Lux.σ), Lux.Dense(n, n, Lux.σ), Lux.Dense(n, 1)) #Neural network from OptimizationFlux library
indvars = [t, x1, x2, x3, x4] #phisically independent variables
depvars = [u(t, x1, x2, x3, x4)] #dependent (target) variable
dim = length(domains)
losses = []
error = []
times = []
dx_err = 0.2
error_strategy = GridTraining(dx_err)
discretization_ = PhysicsInformedNN(chain, error_strategy)
@named pde_system_ = PDESystem(eq, bcs, domains, indvars, depvars)
prob_ = discretize(pde_system_, discretization_)
function loss_function_(θ, p)
return prob_.f.f(θ, nothing)
end
cb_ = function (p, l)
deltaT_s = time_ns() #Start a clock when the callback begins, this will evaluate questo misurerà anche il calcolo degli uniform error
ctime = time_ns() - startTime - timeCounter #This variable is the time to use for the time benchmark plot
append!(times, ctime / 10^9) #Conversion nanosec to seconds
append!(losses, l)
loss_ = loss_function_(p, nothing)
append!(error, loss_)
timeCounter = timeCounter + time_ns() - deltaT_s #timeCounter sums all delays due to the callback functions of the previous iterations
#if (ctime/10^9 > time) #if I exceed the limit time I stop the training
# return true #Stop the minimizer and continue from line 142
#end
return false
end
@named pde_system = PDESystem(eq, bcs, domains, indvars, depvars)
discretization = NeuralPDE.PhysicsInformedNN(chain, strategy)
prob = NeuralPDE.discretize(pde_system, discretization)
timeCounter = 0.0
startTime = time_ns() #Fix initial time (t=0) before starting the training
res = Optimization.solve(prob, minimizer, callback=cb_, maxiters=maxIters)
phi = discretization.phi
params = res.minimizer
# Model prediction
domain = [ts, x1s, x2s, x3s, x4s]
u_predict = [reshape([first(phi([t, x1, x2, x3, x4], res.minimizer; device=cdev)) for x1 in x1s for x2 in x2s for x3 in x3s for x4 in x4s], (length(x1s), length(x2s), length(x3s), length(x4s))) for t in ts] #matrix of model's prediction
return [error, params, domain, times, losses]
endInitially the plot looked really wonky and that was because the loss function wasn't being calculated correctly. So I picked up the loss function from the initial implementation which was already present in the SciMLBenchmarks.jl documentation. So the error occurred when I added the following line and made changes accordingly : function loss_function_(θ, p)
return prob_.f.f(θ, nothing)
end The error is this : MethodError: no method matching (::MLDataDevices.UnknownDevice)(::Matrix{Float64})
Stacktrace:
[1] (::NeuralPDE.Phi{StatefulLuxLayer{Static.True, Chain{@NamedTuple{layer_1::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_2::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_3::Dense{typeof(identity), Int64, Int64, Nothing, Nothing, Static.True}}, Nothing}, Nothing, @NamedTuple{layer_1::@NamedTuple{}, layer_2::@NamedTuple{}, layer_3::@NamedTuple{}}}})(x::Matrix{Float64}, θ::Optimization.OptimizationState{ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Float64, ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Nothing, Optimisers.Leaf{Optimisers.Adam, Tuple{Vector{Float64}, Vector{Float64}, Tuple{Float64, Float64}}}})
@ NeuralPDE C:\Users\Hp\.julia\packages\NeuralPDE\nYBAW\src\pinn_types.jl:42
[2] (::NeuralPDE.var"#7#8")(cord::Matrix{Float64}, θ::Optimization.OptimizationState{ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Float64, ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Nothing, Optimisers.Leaf{Optimisers.Adam, Tuple{Vector{Float64}, Vector{Float64}, Tuple{Float64, Float64}}}}, phi::NeuralPDE.Phi{StatefulLuxLayer{Static.True, Chain{@NamedTuple{layer_1::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_2::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_3::Dense{typeof(identity), Int64, Int64, Nothing, Nothing, Static.True}}, Nothing}, Nothing, @NamedTuple{layer_1::@NamedTuple{}, layer_2::@NamedTuple{}, layer_3::@NamedTuple{}}}})
@ NeuralPDE C:\Users\Hp\.julia\packages\NeuralPDE\nYBAW\src\pinn_types.jl:354
[3] numeric_derivative(phi::NeuralPDE.Phi{StatefulLuxLayer{Static.True, Chain{@NamedTuple{layer_1::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_2::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_3::Dense{typeof(identity), Int64, Int64, Nothing, Nothing, Static.True}}, Nothing}, Nothing, @NamedTuple{layer_1::@NamedTuple{}, layer_2::@NamedTuple{}, layer_3::@NamedTuple{}}}}, u::NeuralPDE.var"#7#8", x::Matrix{Float64}, εs::Vector{Vector{Float64}}, order::Int64, θ::Optimization.OptimizationState{ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Float64, ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Nothing, Optimisers.Leaf{Optimisers.Adam, Tuple{Vector{Float64}, Vector{Float64}, Tuple{Float64, Float64}}}})
@ NeuralPDE C:\Users\Hp\.julia\packages\NeuralPDE\nYBAW\src\pinn_types.jl:382
[4] macro expansion
@ C:\Users\Hp\.julia\packages\NeuralPDE\nYBAW\src\discretize.jl:130 [inlined]
[5] macro expansion
@ C:\Users\Hp\.julia\packages\RuntimeGeneratedFunctions\M9ZX8\src\RuntimeGeneratedFunctions.jl:163 [inlined]
[6] macro expansion
@ .\none:0 [inlined]
[7] generated_callfunc(::RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:cord, Symbol("##θ#226"), :phi, :derivative, :integral, :u, :p), NeuralPDE.var"#_RGF_ModTag", NeuralPDE.var"#_RGF_ModTag", (0x875c3d6c, 0x2ff644f8, 0x57c4854c, 0xf710f944, 0x14a70865), Expr}, ::Matrix{Float64}, ::Optimization.OptimizationState{ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Float64, ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Nothing, Optimisers.Leaf{Optimisers.Adam, Tuple{Vector{Float64}, Vector{Float64}, Tuple{Float64, Float64}}}}, ::NeuralPDE.Phi{StatefulLuxLayer{Static.True, Chain{@NamedTuple{layer_1::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_2::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_3::Dense{typeof(identity), Int64, Int64, Nothing, Nothing, Static.True}}, Nothing}, Nothing, @NamedTuple{layer_1::@NamedTuple{}, layer_2::@NamedTuple{}, layer_3::@NamedTuple{}}}}, ::typeof(NeuralPDE.numeric_derivative), ::NeuralPDE.var"#239#246"{NeuralPDE.var"#239#240#247"{typeof(NeuralPDE.numeric_derivative)}, Dict{Symbol, Int64}, Dict{Symbol, Int64}, GridTraining{Float64}}, ::NeuralPDE.var"#7#8", ::Nothing)
@ NeuralPDE .\none:0
[8] (::RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:cord, Symbol("##θ#226"), :phi, :derivative, :integral, :u, :p), NeuralPDE.var"#_RGF_ModTag", NeuralPDE.var"#_RGF_ModTag", (0x875c3d6c, 0x2ff644f8, 0x57c4854c, 0xf710f944, 0x14a70865), Expr})(::Matrix{Float64}, ::Optimization.OptimizationState{ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Float64, ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Nothing, Optimisers.Leaf{Optimisers.Adam, Tuple{Vector{Float64}, Vector{Float64}, Tuple{Float64, Float64}}}}, ::NeuralPDE.Phi{StatefulLuxLayer{Static.True, Chain{@NamedTuple{layer_1::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_2::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_3::Dense{typeof(identity), Int64, Int64, Nothing, Nothing, Static.True}}, Nothing}, Nothing, @NamedTuple{layer_1::@NamedTuple{}, layer_2::@NamedTuple{}, layer_3::@NamedTuple{}}}}, ::Function, ::Function, ::Function, ::Nothing)
@ RuntimeGeneratedFunctions C:\Users\Hp\.julia\packages\RuntimeGeneratedFunctions\M9ZX8\src\RuntimeGeneratedFunctions.jl:150
[9] (::NeuralPDE.var"#197#198"{RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:cord, Symbol("##θ#226"), :phi, :derivative, :integral, :u, :p), NeuralPDE.var"#_RGF_ModTag", NeuralPDE.var"#_RGF_ModTag", (0x875c3d6c, 0x2ff644f8, 0x57c4854c, 0xf710f944, 0x14a70865), Expr}, NeuralPDE.var"#7#8", NeuralPDE.var"#239#246"{NeuralPDE.var"#239#240#247"{typeof(NeuralPDE.numeric_derivative)}, Dict{Symbol, Int64}, Dict{Symbol, Int64}, GridTraining{Float64}}, typeof(NeuralPDE.numeric_derivative), NeuralPDE.Phi{StatefulLuxLayer{Static.True, Chain{@NamedTuple{layer_1::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_2::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_3::Dense{typeof(identity), Int64, Int64, Nothing, Nothing, Static.True}}, Nothing}, Nothing, @NamedTuple{layer_1::@NamedTuple{}, layer_2::@NamedTuple{}, layer_3::@NamedTuple{}}}}, Nothing})(cord::Matrix{Float64}, θ::Optimization.OptimizationState{ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Float64, ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Nothing, Optimisers.Leaf{Optimisers.Adam, Tuple{Vector{Float64}, Vector{Float64}, Tuple{Float64, Float64}}}})
@ NeuralPDE C:\Users\Hp\.julia\packages\NeuralPDE\nYBAW\src\discretize.jl:150
[10] (::NeuralPDE.var"#78#79"{NeuralPDE.var"#197#198"{RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:cord, Symbol("##θ#226"), :phi, :derivative, :integral, :u, :p), NeuralPDE.var"#_RGF_ModTag", NeuralPDE.var"#_RGF_ModTag", (0x875c3d6c, 0x2ff644f8, 0x57c4854c, 0xf710f944, 0x14a70865), Expr}, NeuralPDE.var"#7#8", NeuralPDE.var"#239#246"{NeuralPDE.var"#239#240#247"{typeof(NeuralPDE.numeric_derivative)}, Dict{Symbol, Int64}, Dict{Symbol, Int64}, GridTraining{Float64}}, typeof(NeuralPDE.numeric_derivative), NeuralPDE.Phi{StatefulLuxLayer{Static.True, Chain{@NamedTuple{layer_1::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_2::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_3::Dense{typeof(identity), Int64, Int64, Nothing, Nothing, Static.True}}, Nothing}, Nothing, @NamedTuple{layer_1::@NamedTuple{}, layer_2::@NamedTuple{}, layer_3::@NamedTuple{}}}}, Nothing}, Matrix{Float64}})(θ::Optimization.OptimizationState{ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Float64, ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Nothing, Optimisers.Leaf{Optimisers.Adam, Tuple{Vector{Float64}, Vector{Float64}, Tuple{Float64, Float64}}}})
@ NeuralPDE C:\Users\Hp\.julia\packages\NeuralPDE\nYBAW\src\training_strategies.jl:70
[11] (::NeuralPDE.var"#263#284"{Optimization.OptimizationState{ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Float64, ComponentArrays.ComponentVector{Float64, Vector{Float64}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:60, Axis(weight = ViewAxis(1:50, ShapedAxis((10, 5))), bias = ViewAxis(51:60, Shaped1DAxis((10,))))), layer_2 = ViewAxis(61:170, Axis(weight = ViewAxis(1:100, ShapedAxis((10, 10))), bias = ViewAxis(101:110, Shaped1DAxis((10,))))), layer_3 = ViewAxis(171:181, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10))), bias = ViewAxis(11:11, Shaped1DAxis((1,))))))}}}, Nothing, Optimisers.Leaf{Optimisers.Adam, Tuple{Vector{Float64}, Vector{Float64}, Tuple{Float64, Float64}}}}})(pde_loss_function::NeuralPDE.var"#78#79"{NeuralPDE.var"#197#198"{RuntimeGeneratedFunctions.RuntimeGeneratedFunction{(:cord, Symbol("##θ#226"), :phi, :derivative, :integral, :u, :p), NeuralPDE.var"#_RGF_ModTag", NeuralPDE.var"#_RGF_ModTag", (0x875c3d6c, 0x2ff644f8, 0x57c4854c, 0xf710f944, 0x14a70865), Expr}, NeuralPDE.var"#7#8", NeuralPDE.var"#239#246"{NeuralPDE.var"#239#240#247"{typeof(NeuralPDE.numeric_derivative)}, Dict{Symbol, Int64}, Dict{Symbol, Int64}, GridTraining{Float64}}, typeof(NeuralPDE.numeric_derivative), NeuralPDE.Phi{StatefulLuxLayer{Static.True, Chain{@NamedTuple{layer_1::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_2::Dense{typeof(σ), Int64, Int64, Nothing, Nothing, Static.True}, layer_3::Dense{typeof(identity), Int64, Int64, Nothing, Nothing, Static.True}}, Nothing}, Nothing, @NamedTuple{layer_1::@NamedTuple{}, layer_2::@NamedTuple{}, layer_3::@NamedTuple{}}}}, Nothing}, Matrix{Float64}})
@ NeuralPDE .\none:0
...
@ SciMLBase C:\Users\Hp\.julia\packages\SciMLBase\sYmAV\src\solve.jl:95
[22] allen_cahn(strategy::QuadratureTraining{Float64, CubaCuhre}, minimizer::Optimisers.Adam, maxIters::Int64)
@ Main e:\SciMLBenchmarks.jl\benchmarks\PINNErrorsVsTime\jl_notebook_cell_df34fa98e69747e1a8f8a730347b8e2f_W0sZmlsZQ==.jl:116
[23] top-level scope
@ e:\SciMLBenchmarks.jl\benchmarks\PINNErrorsVsTime\jl_notebook_cell_df34fa98e69747e1a8f8a730347b8e2f_W2sZmlsZQ==.jl:4Which tells that there's some method error with the matrix. The stacktrace suggests that there is some problem with the Phi layer in NeuralPDE.jl specifically this seems to be coming from pinn_types.jl line 42. But not sure what changes are needed in order to incorporate this |
|
Try calling the loss function like |
Yeah, this worked. Thank you so much. Also curious about this thing, is there any reference for this, where I can learn this from. This was not that easy to observe for me 😅 |
|
This is the graph generated this time |
|
I removed the HCubatureJL algorithm due to its incompatibility with NeuralPDE.jl |
|
@ChrisRackauckas all the CI checks here have passed. Can you please review this ?? |
|
Can you post how the plot looks like for alan_cahn now? |
|
@ChrisRackauckas @sathvikbhagavan Does this graph still needs some changes ? |
|
Shouldn't all strategies start at t=0? Also plot time in log |
|
Just about, yes. Move the legend to the outside |
|
@ChrisRackauckas @sathvikbhagavan I incorporated the suggested changes. I haven't pushed the code yet but this is how the plot looks like when I ran this code locally |
|
@ChrisRackauckas @sathvikbhagavan does this graph look satisfactory or any changes are needed to this ? |
|
I haven't pushed the code yet because once I do so it would take 3 days for the CI to complete and for us to get the results. It takes only 7 minutes on my local system so if any changes are needed I can do it right away and we can see the results too, |
|
A couple of things can be done to improve the plot:
|
|
The weird results seem to be a consequence of limitations of the quadrature libraries. The quadrature libraries for the cubatures have minimum numbers of steps they allow, and then have growth by odd factors. In previous plots they had failed, now they work because they clamp the iters to an allowed quantity, such as N^3 for 3D. This might mean there's only value that is in the allowed range, and then the next one takes way to long to compute. So, this is okay, for odd reasons. |
|
@ParamThakkar123 thanks! Go forth and claim your reward. You did well here. It took awhile for me to understand why the results were like this, and I couldn't find you did anything wrong here, it's just interesting behavior of the integration libraries. |
ChrisRackauckas
added a commit
to SciML/sciml.ai
that referenced
this pull request
May 6, 2025
This PR certifies the completion of the two bnechmark sets by @ParamThakkar123 with the following PRs: * SciML/SciMLBenchmarks.jl#1148 * SciML/SciMLBenchmarks.jl#1160 * SciML/SciMLBenchmarks.jl#1159 It also adds two new benchmark sets that are in dire need of updates. As a monetary decision, requires 3 approvals. I approve.
|
Hello @ChrisRackauckas, really glad that the things worked out and the PR got merged. Actually looking at the results even I was quite puzzled and got into thinking that something is wrong here and throughout this I spent a lot of time into getting them right. Glad that we reached a conclusion. It' really weird and interesting too that we got to see these results |
|
Thanks for merging my PR. I will start a payout request on open collective to claim the reward |
|
Always happy to contribute! Please do let me know if you have any more projects where I can contribute my best. |
|
@ChrisRackauckas What amount should be the amount of the reward that I should set in the Open collective invoice ? |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Checklist
contributor guidelines, in particular the SciML Style Guide and
COLPRAC.
Additional context
Add any other context about the problem here.