d9w / cartesiangeneticprogramming.jl Goto Github PK

This may be an issue in PyCall.jl. Normally env[:render] returns true and displays the environment (for example, for env = gym.make("CartPole-v0"), but for pybullet envs this doesn't work:

julia> env[:render](mode="human")
0-element Array{Float64,1}

Add parallel evaluation to oneplus, GA

julia experiments/atari.jl --id boxing

WARNING: Method definition info(Any...) in module Base at util.jl:534 overwritten in module Logging at /home/mostafa/.julia/v0.6/Logging/src/Logging.jl:115.
WARNING: Method definition warn(Any...) in module Base at util.jl:589 overwritten in module Logging at /home/mostafa/.julia/v0.6/Logging/src/Logging.jl:115.
ERROR: LoadError: UndefVarError: Game not defined
Stacktrace:
[1] get_params(::Dict{String,Any}) at /home/mostafa/CGP.jl/experiments/atari.jl:71
[2] include_from_node1(::String) at ./loading.jl:576
[3] include(::String) at ./sysimg.jl:14
[4] process_options(::Base.JLOptions) at ./client.jl:305
[5] _start() at ./client.jl:371
while loading /home/mostafa/CGP.jl/experiments/atari.jl, in expression starting on line 113

aside:
i already tried that #8

MicroLogging.jl will be used in 0.7, also fixes the warnings from Logging.jl:

WARNING: Method definition warn(Any...) in module Base at util.jl:589 overwritten in module Logging at /home/d9w/.julia/v0.6/Logging/src/Logging.jl:115.

Config as a submodule introduces a lot of complications. A mutable struct might be sufficient, if it doesn't drastically decrease performance. A dict could also work.

In relation to the Atari experiment, I'm trying to understand the structure of rgenes in cgp.jl. I hope that I'm looking at right file for that experiment.

I thought the 1st element of rgenes is the function. The only line that looks relevant to me is https://github.com/d9w/CGP.jl/blob/master/src/chromosomes/cgp.jl#L34 but that's using rgenes[:, 3]. I guess my question is: should that be rgenes[:, 1], or have I misunderstood?

I see at https://github.com/d9w/CGP.jl/blob/master/src/chromosomes/cgp.jl#L37 that the parameter is the 4th element. Makes perfect sense to me. And it looks like the 2nd and 3rd element are the x, y node inputs at https://github.com/d9w/CGP.jl/blob/master/src/chromosomes/cgp.jl#L19

Thank you for your help.

I'm new to Julia so pardon me if the questions are n00b, but here's the issue:

Problem

1. Does not include "PaddedViewer" in depencency (fixed by manually changing deps in environment)
2. running julia dev/CGP/experiments/xor.jl, get back

ERROR: LoadError: MethodError: Cannot `convert` an object of type Nothing to an object of type Union{UUID, Bool}
Closest candidates are:
  convert(::Type{T}, !Matched::T) where T at essentials.jl:154
Stacktrace:
 [1] explicit_manifest_deps_get(::String, ::Base.UUID, ::String) at ./loading.jl:478
 [2] manifest_deps_get(::String, ::Base.PkgId, ::String) at ./loading.jl:307
 [3] identify_package(::Base.PkgId, ::String) at ./loading.jl:203
 [4] identify_package at ./loading.jl:196 [inlined]
 [5] require(::Module, ::Symbol) at ./loading.jl:812
 [6] include at ./boot.jl:317 [inlined]
 [7] include_relative(::Module, ::String) at ./loading.jl:1038
 [8] include(::Module, ::String) at ./sysimg.jl:29
 [9] top-level scope at none:2
 [10] eval at ./boot.jl:319 [inlined]
 [11] eval(::Expr) at ./client.jl:389
 [12] top-level scope at ./none:3
in expression starting at /home/tingkailiu/.julia/dev/CGP/src/CGP.jl:4
ERROR: LoadError: Failed to precompile CGP [4fd9c4f4-07f5-549d-bd68-3a67ae5bbbb9] to /home/tingkailiu/.julia/compiled/v1.0/CGP/0AjpJ.ji.
Stacktrace:
 [1] error(::String) at ./error.jl:33
 [2] macro expansion at ./logging.jl:313 [inlined]
 [3] compilecache(::Base.PkgId, ::String) at ./loading.jl:1184
 [4] _require(::Base.PkgId) at ./logging.jl:311
 [5] require(::Base.PkgId) at ./loading.jl:852
 [6] macro expansion at ./logging.jl:311 [inlined]
 [7] require(::Module, ::Symbol) at ./loading.jl:834
 [8] include at ./boot.jl:317 [inlined]
 [9] include_relative(::Module, ::String) at ./loading.jl:1038
 [10] include(::Module, ::String) at ./sysimg.jl:29
 [11] exec_options(::Base.JLOptions) at ./client.jl:229
 [12] _start() at ./client.jl:421
in expression starting at /home/tingkailiu/.julia/dev/CGP/experiments/xor.jl:1

Thanks for open-sourcing this code, been meaning to trying out the algorithms in the recent Atari paper for a while.

add weights to nodes, activation functions etc?

http://www.cartesiangp.co.uk/resources.html gives me "404 Not Found".

I wanted some way to visualize the graph and noticed graphs/graph_utils.jl which seems to have been the way graphs were visualized in the past. However, it seems to be out of date.

Had to add: using Printf in order for @sprintf to work.

Also notice that Chromosome is not defined anywhere:
ERROR: LoadError: UndefVarError: Chromosome not defined

Hello - thank you for sharing this project: very cool!

This issue is almost certainly my problem, but pointers would be appreciated.

I've worked through the examples in the readme (xor, etc) and they all work great. For the Atari experiments I'm hitting:

$ julia experiments/atari.jl
WARNING: Method definition info(Any...) in module Base at util.jl:534 overwritten in module Logging at ~/julia/v0.6/Logging/src/Logging.jl:115.
WARNING: Method definition warn(Any...) in module Base at util.jl:589 overwritten in module Logging at ~/.julia/v0.6/Logging/src/Logging.jl:115.
ERROR: LoadError: UndefVarError: Game not defined
Stacktrace:
 [1] get_params(::Dict{String,Any}) at CGP.jl/experiments/atari.jl:71
 [2] include_from_node1(::String) at ./loading.jl:576
 [3] include(::String) at ./sysimg.jl:14
 [4] process_options(::Base.JLOptions) at ./client.jl:305
 [5] _start() at ./client.jl:371

I'm on MacOS, I have Julia 0.6.4 installed, and I've run:

Pkg.add("ArcadeLearningEnvironment")

I assume I'm missing another step.

In the Atari experiment there's a probablistic action repeat (frame skip?):

https://github.com/d9w/CGP.jl/blob/d82cc4b48c9c957a971bddea8f8da916824f231f/experiments/atari.jl#L23-L25

My reading of this is that the behaviour (or something very similar) is enabled by default in ALE (since 0.5):

When you call self.ale.act(a) for some action a, the action a will be executed right away
with a probability (1 - p), otherwise, the action that was previously
executed will be executed again. The default value for p right now is 0.25.

^^ from: Farama-Foundation/Arcade-Learning-Environment#166 (comment)

I assume it's safe to remove this test from the code (unless there's an ale.cfg in place to disable the default ALE behaviour).

Apologies in advance if I've mised the point :-/ ...still learning.

• names are truncated based on final "_", should be first
• show x, y inputs in different colors
• make self-references a loop

Some kind of evolution algorithm:
https://github.com/S6Regen/Thunderbird
There is a paper on the internet "Continuous Gray code optimization" that describes the algorithm used.