autopush

Author: SCiarella

Project.toml

@@ -36,6 +36,7 @@ OptimizationOptimJL = "36348300-93cb-4f02-beb5-3c3902f8871e"
 ParameterSchedulers = "d7d3b36b-41b8-4d0d-a2bf-768c6151755e"
 Printf = "de0858da-6303-5e67-8744-51eddeeeb8d7"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
+SciMLSensitivity = "1ed8b502-d754-442c-8d5d-10ac956f44a1"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 
 [weakdeps]
@@ -78,6 +79,7 @@ Optimisers = "0.4"
 OptimizationCMAEvolutionStrategy = "0.3.0"
 OptimizationOptimJL = "0.4.3"
 ParameterSchedulers = "0.4"
+SciMLSensitivity = "7.84.0"
 Statistics = "1.11.1"
 cuDNN = "1"
 julia = "1.11"

benchmark.jl

@@ -18,8 +18,7 @@ ispath(compdir) || mkpath(compdir)
 
 # List configurations files
 using Glob
-exclude_patterns = ["att", "cno"]
-exclude_patterns = ["cno"]
+exclude_patterns = ["att", "cno", "cnn_ins", "_1"]
 @warn "Excluding configurations with patterns: $(exclude_patterns)"
 all_confs = glob("*.yaml", confdir)
 list_confs = filter(conf -> all(!occursin(pat, conf) for pat in exclude_patterns), all_confs)

cnn_model_workflow.jl

@@ -106,6 +106,7 @@ using OptimizationOptimJL
 using OptimizationCMAEvolutionStrategy
 using ParameterSchedulers
 using Random
+using SciMLSensitivity
 
 
 # ## Random number seeds
@@ -350,6 +351,19 @@ projectorders = eval(Meta.parse(conf["posteriori"]["projectorders"]))
 nprojectorders = length(projectorders)
 @assert nprojectorders == 1 "Only DCF should be done"
 
+sensealg = haskey(conf["posteriori"], "sensealg") ? eval(Meta.parse(conf["posteriori"]["sensealg"])) : nothing
+sciml_solver = haskey(conf["posteriori"], "sciml_solver") ? eval(Meta.parse(conf["posteriori"]["sciml_solver"])) : nothing
+if sensealg !== nothing
+    @info "Using sensitivity algorithm: $sensealg"
+else
+    @info "No sensitivity algorithm specified"
+end
+if sciml_solver !== nothing
+    @info "Using SciML solver: $sciml_solver"
+else
+    @info "No SciML solver specified"
+end
+
 # Train
 for i = 1:ntrajectory
 	if i%numtasks == taskid -1
@@ -377,7 +391,8 @@ let
         nepoch,
         do_plot = conf["posteriori"]["do_plot"],
         plot_train = conf["posteriori"]["plot_train"],
-        sensealg = haskey(conf["posteriori"],:sensealg) ? eval(Meta.parse(conf["posteriori"]["sensealg"])) : nothing,
+        sensealg = sensealg,
+        sciml_solver = sciml_solver,
         dataproj = conf["dataproj"],
     )
 end
@@ -522,14 +537,14 @@ let
         dudt_nomod = NS.create_right_hand_side_inplace(
             setup, psolver)
 
-        epost.nomodel[I,:], _ = compute_epost(dudt_nomod, θ_cnn_post[I].*0 , tspan, data, tsave, dt)
+        epost.nomodel[I,:], _ = compute_epost(dudt_nomod, sciml_solver, θ_cnn_post[I].*0 , tspan, data, tsave, dt)
         @info "Epost nomodel" epost.nomodel[I,:]
         # with closure
         dudt = NS.create_right_hand_side_with_closure_inplace(
             setup, psolver, closure, st)
-        epost.model_prior[I, :], _ = compute_epost(dudt, device(θ_cnn_prior[ig, ifil]) , tspan, data, tsave, dt)
+        epost.model_prior[I, :], _ = compute_epost(dudt, sciml_solver, device(θ_cnn_prior[ig, ifil]) , tspan, data, tsave, dt)
         @info "Epost model_prior" epost.model_prior[I, :]
-        epost.model_post[I, :], epost.model_t_post_inference[I] = compute_epost(dudt, device(θ_cnn_post[ig, ifil, iorder]) , tspan, data, tsave, dt)
+        epost.model_post[I, :], epost.model_t_post_inference[I] = compute_epost(dudt, sciml_solver, device(θ_cnn_post[ig, ifil, iorder]) , tspan, data, tsave, dt)
         @info "Epost model_post" epost.model_post[I, :]
 
         clean()

configs/snellius/cnn_1.yaml

@@ -50,3 +50,4 @@ posteriori:
   dt: 0.0001
   do_plot: false
   plot_train: false
+  sciml_solver: "Tsit5()"

configs/snellius/cnn_CMA.yaml

@@ -51,4 +51,4 @@ posteriori:
   dt: 0.0001
   do_plot: false
   plot_train: false
-  sensealg: "InterpolatingAdjoint()"
+  sciml_solver: "Tsit5()"

configs/snellius/cnn_backsol.yaml

@@ -2,7 +2,7 @@ docreatedata: true
 docomp: true
 ntrajectory: 8
 T: "Float32"
-dataproj: true
+dataproj: false 
 params:
   D: 2
   lims: [0.0, 1.0]
@@ -33,7 +33,7 @@ closure:
   use_bias: [true, true, true, true, false]
   rng: "Xoshiro(seeds.θ_start)"
 priori:
-  reuse: "cnn_project"
+  reuse: "cnn_noproj"
   dotrain: true
   nepoch: 50000
   batchsize: 64
@@ -53,3 +53,4 @@ posteriori:
   do_plot: false
   plot_train: false
   sensealg: "BacksolveAdjoint()"
+  sciml_solver: "Tsit5()"

configs/snellius/cnn_ins.yaml

@@ -50,3 +50,4 @@ posteriori:
   do_plot: false
   plot_train: false
   nsamples: 1
+  sciml_solver: "Tsit5()"

configs/snellius/cnn_interp.yaml

@@ -2,7 +2,7 @@ docreatedata: true
 docomp: true
 ntrajectory: 8
 T: "Float32"
-dataproj: true
+dataproj: false
 params:
   D: 2
   lims: [0.0, 1.0]
@@ -33,7 +33,7 @@ closure:
   use_bias: [true, true, true, true, false]
   rng: "Xoshiro(seeds.θ_start)"
 priori:
-  reuse: "cnn_project"
+  reuse: "cnn_noproj"
   dotrain: true
   nepoch: 50000
   batchsize: 64
@@ -51,4 +51,6 @@ posteriori:
   dt: 0.0001
   do_plot: false
   plot_train: false
+  #sensealg: "InterpolatingAdjoint(autojacvec=EnzymeVJP())"
   sensealg: "InterpolatingAdjoint()"
+  sciml_solver: "Tsit5()"

configs/snellius/cnn_lbfgs.yaml

@@ -53,3 +53,4 @@ posteriori:
   do_plot: false
   plot_train: false
   sensealg: "InterpolatingAdjoint()"
+  sciml_solver: "Tsit5()"

configs/snellius/cnn_rk4.yaml

@@ -0,0 +1,54 @@
+docreatedata: true
+docomp: true
+ntrajectory: 8
+T: "Float32"
+dataproj: false
+params:
+  D: 2
+  lims: [0.0, 1.0]
+  Re: 6000.0
+  tburn: 0.5
+  tsim: 5.0
+  savefreq: 50
+  ndns: 4096
+  nles: [64]
+  filters: ["FaceAverage()"]
+  icfunc: "(setup, psolver, rng) -> random_field(setup, T(0); kp=20, psolver, rng)"
+  method: "RKMethods.Wray3(; T)"
+  bodyforce: "(dim, x, y, t) -> (dim == 1) * 5 * sinpi(8 * y)"
+  issteadybodyforce: true
+  processors: "(; log = timelogger(; nupdate=100))"
+  Δt: 0.00005
+seeds:
+  dns: 123456
+  θ_start: 234
+  prior: 345
+  post: 456
+closure:
+  name: "cnn_noproj"
+  type: cnn
+  radii: [2, 2, 2, 2, 2]
+  channels: [24, 24, 24, 24, 2]
+  activations: ["tanh", "tanh", "tanh", "tanh", "identity"]
+  use_bias: [true, true, true, true, false]
+  rng: "Xoshiro(seeds.θ_start)"
+priori:
+  reuse: "cnn_noproj"
+  dotrain: true
+  nepoch: 50000
+  batchsize: 64
+  opt: "OptimiserChain(Adam(T(1.0e-3)), ClipGrad(0.1))"
+  do_plot: false
+  plot_train: false
+posteriori:
+  dotrain: true
+  projectorders: "(ProjectOrder.Last, )"
+  nepoch: 1500
+  opt: "OptimiserChain(Adam(T(1.0e-3)), ClipGrad(0.1))"
+  nunroll: 5
+  nunroll_valid: 10
+  dt: 0.0001
+  do_plot: false
+  plot_train: false
+  nsamples: 1
+  sciml_solver: "RK4()"