axsk

julia> JET.@report_opt u([1.],1.)
═════ 77 possible errors found ═════
┌ @ /home/htc/bzfsikor/code/OptImpSampling.jl/src/control.jl:137 dlogz = fgrad(#148, x)
│┌ @ /home/htc/bzfsikor/code/OptImpSampling.jl/src/ociso.jl:25 ForwardDiff.gradient(f, x)
││┌ @ /home/htc/bzfsikor/.julia/packages/ForwardDiff/pDtsf/src/gradient.jl:17 ForwardDiff.GradientConfig(f, x)
│││┌ @ /home/htc/bzfsikor/.julia/packages/ForwardDiff/pDtsf/src/config.jl:121 ForwardDiff.Chunk(x)
││││┌ @ /home/htc/bzfsikor/.julia/packages/ForwardDiff/pDtsf/src/prelude.jl:26 #self#(x, ForwardDiff.DEFAULT_CHUNK_THRESHOLD)
│││││┌ @ /home/htc/bzfsikor/.julia/packages/ForwardDiff/pDtsf/src/prelude.jl:26 ForwardDiff.Chunk(ForwardDiff.length(x), threshold)
││││││┌ @ /home/htc/bzfsikor/.julia/packages/ForwardDiff/pDtsf/src/prelude.jl:22 %13()
│││││││ runtime dispatch detected: %13::Type{ForwardDiff.Chunk{_A}} where _A()::ForwardDiff.Chunk

axsk

# basically a type stable version of memoize.jl with a 1-element dict/cache
# it would be nice however to be able to only call cachedf = cache(f)
# such that cachedf would be typle stable.
# so far i have no idea how to achieve this

mutable struct CachedFunction{F, I, O}
    f::F
    in::I
    out::O
end

axsk

#https://github.com/FluxML/Zygote.jl/issues/1294

using Zygote
using StochasticDiffEq, SciMLSensitivity
using StatsBase
import Lux

function mwe()
    x0 = rand(1)
    p0 = rand(1)

axsk

# https://github.com/SciML/SciMLSensitivity.jl/issues/720
using Zygote
using StochasticDiffEq, SciMLSensitivity

function mwe1()
    x0 = [0.]
    drift(dx, x, p, t) = (dx .= p)
    noise(dx, x, p, t) = (dx .= 0.)
    n0 = zeros(1,1)
    T = 100.

axsk

ERROR: Event was triggered but no corresponding event in ContinuousCallback was found. Please report this error.
Stacktrace:
  [1] error(s::String)
    @ Base ./error.jl:35
  [2] get_indx(cb::ContinuousCallback{var"#9#10"{Int64, Int64}, SciMLSensitivity.TrackedAffect{Float64, Vector{Float64}, ComponentArrays.ComponentVector{Float32, Vector{Float32}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:20, Axis(weight = ViewAxis(1:10, ShapedAxis((10, 1), NamedTuple())), bias = ViewAxis(11:20, ShapedAxis((10, 1), NamedTuple())))), layer_2 = ViewAxis(21:31, Axis(weight = ViewAxis(1:10, ShapedAxis((1, 10), NamedTuple())), bias = ViewAxis(11:11, ShapedAxis((1, 1), NamedTuple())))))}}}, typeof(terminate!), SciMLSensitivity.ImplicitCorrection{Vector{Float64}, Vector{Float64}, SciMLSensitivity.ConditionTimeWrapper{var"#9#10"{Int64, Int64}, Vector{Float64}, SciMLSensitivity.FakeIntegrator{Vector{Float64}, ComponentArrays.ComponentVector{Float32, Vector{Float32}, Tuple{ComponentArrays.Axis{(layer_1 = ViewAxis(1:20, Axis(we

axsk

import numpy as np

t_adopt = .5
p_adopt = .9
p_expl = .1

def timestep(state, influence):
    influenced = (np.abs(np.dot(influence,state)) > t_adopt)
    r = np.random.rand(np.size(state))
    # vectorized: 415us

axsk

using Sundials

function test1()
  yS0 = [1. 2. 3.; 4. 5 6]
  N, Ns = size(yS0)

  GC.@preserve yS0n = [Sundials.NVector(yS0[:,j]) for j=1:Ns] # array of julia wrappers
  GC.@preserve yS0nv = [Sundials.N_Vector(n) for n in yS0n]   # array of pointers to sund. structs
  pyS0  = pointer(yS0nv)

axsk

using Sundials
using Sundials: N_Vector, N_Vector_S

using ForwardDiff
#using ReverseDiff

"""
    sens(f!, t0, y0, p, tout; reltol, abstol)

Compute the solution and sensivities to the parametrized ODE problem defined by `f!(ẏ, t, y, p)`, starting at t0, y0, p.

axsk

using Plots

rect(x,y,w,h) = Shape(x+[0,w,w,0], y+[0,0,h,h])

@recipe function f(ds::Array{Measurement}; alpha = 1)
  isempty(ds) && return []
  binedges = [0,10,12,14,16,18,20,25]
  nmax = nfollicles(ds)
  a = alpha

axsk

heu