Hyperopt.jl
Hyperopt.jl copied to clipboard
baggepinnen
Hyperparameter optimization in Julia.
I'm using Random Sampler, and thought if gave the number of iteration equal to the length of brute-force size, it would do it automatically. However, I noticed that the sampler...
phyperopt seems to be broken MWE: ```julia using Distributed addprocs(4; exeflags=`--project=$(Base.active_project())`) using Hyperopt @everywhere using Optim @everywhere f(a;c=10) = sum(@. 100 + (a-3)^2 + (c-100)^2) hohb = @phyperopt for resources=50,...
In citing this package, I get the following warning: ``` julia> using PkgCite julia> get_tool_citation(texttt = true, badge = true) ┌ Warning: Entry bagge2018hyperopt is missing the journal field(s). └...
Currently, baysean estimates seem to take on any real value. This is a problem when a hyperparameter represents e.g. a probability. See for example how the sampled distribution ends up...
The problem arises when doing BOHB. This package currently makes no distinction between trainable parameters and hyperparameters, but they are in my mind very separate things. This is closely related...
I am currently getting the following error ``` ERROR: MethodError: no method matching /(::Type{Momentum}, ::Int64) Closest candidates are: /(::Any, ::ChainRulesCore.AbstractThunk) at ~/.julia/packages/ChainRulesCore/RbX5a/src/tangent_types/thunks.jl:33 /(::Union{Int128, Int16, Int32, Int64, Int8, UInt128, UInt16, UInt32,...
What does the first argument to `@hyperopt` do when the sampler is Hyperband? I am unable to see it having any influence on the value of `resources` inside the loop,...
Currently, when doing hyperband, it is allowed to pass on a loss as the first element, and a "state" at the second element. The "state" is what you can use...
I currently have quite a few parameters I am optimizing for. Additionally, I want to run several @hyperopt runs to get some statistics on the quality of the output. My...