initial commit

.travis.yml

@@ -5,7 +5,6 @@ os:
   - osx
 julia:
   - 0.6
-  - nightly
 notifications:
   email: false
 git:

README.md

@@ -1,7 +1,9 @@
 # RLEnvGym
 
-[![Build Status](https://travis-ci.com/Johanni Brea/RLEnvGym.jl.svg?branch=master)](https://travis-ci.com/Johanni Brea/RLEnvGym.jl)
+[![Build Status](https://travis-ci.com/JuliaReinforcementLearning/RLEnvGym.jl.svg?branch=master)](https://travis-ci.com/JuliaReinforcementLearning/RLEnvGym.jl)
 
-[![Coverage Status](https://coveralls.io/repos/Johanni Brea/RLEnvGym.jl/badge.svg?branch=master&service=github)](https://coveralls.io/github/Johanni Brea/RLEnvGym.jl?branch=master)
+[![Coverage Status](https://coveralls.io/repos/JuliaReinforcementLearning/RLEnvGym.jl/badge.svg?branch=master&service=github)](https://coveralls.io/github/JuliaReinforcementLearning/RLEnvGym.jl?branch=master)
 
-[![codecov.io](http://codecov.io/github/Johanni Brea/RLEnvGym.jl/coverage.svg?branch=master)](http://codecov.io/github/Johanni Brea/RLEnvGym.jl?branch=master)
+[![codecov.io](http://codecov.io/github/JuliaReinforcementLearning/RLEnvGym.jl/coverage.svg?branch=master)](http://codecov.io/github/JuliaReinforcementLearning/RLEnvGym.jl?branch=master)
+
+Making the [OpenAI gym](https://github.com/openai/gym) environments available to the [Julia Reinforcement Learning](https://github.com/jbrea/ReinforcementLearning.jl) package.

REQUIRE

@@ -1 +1,4 @@
 julia 0.6
+PyCall
+Reexport
+ReinforcementLearning

deps/build.jl

@@ -0,0 +1,26 @@
+using PyCall
+
+# Change that to whatever packages you need.
+const PACKAGES = ["gym"]
+
+# Use eventual proxy info
+proxy_arg=String[]
+if haskey(ENV, "http_proxy")
+    push!(proxy_arg, "--proxy")
+    push!(proxy_arg, ENV["http_proxy"])
+end
+
+# Import pip
+try
+    @pyimport pip
+catch
+    # If it is not found, install it
+    println("Pip not found on your sytstem. Downloading it.")
+    get_pip = joinpath(dirname(@__FILE__), "get-pip.py")
+    download("https://bootstrap.pypa.io/get-pip.py", get_pip)
+    run(`$(PyCall.python) $(proxy_arg) $get_pip --user`)
+end
+
+println("Installing required python packages using pip")
+run(`$(PyCall.python) $(proxy_arg) -m pip install --user --upgrade pip setuptools`)
+run(`$(PyCall.python) $(proxy_arg) -m pip install --user $(PACKAGES)`)

examples/cartpole.jl

@@ -0,0 +1,25 @@
+using RLEnvGym, Flux
+# List all envs
+
+listallenvs()
+
+# CartPole example
+
+env = GymEnv("CartPole-v0")
+learner = DQN(Chain(Dense(4, 48, relu), Dense(48, 24, relu), Dense(24, 2)),
+                  updateevery = 1, updatetargetevery = 100,
+                  startlearningat = 50, minibatchsize = 32,
+                  doubledqn = false, replaysize = 10^3, 
+                  opttype = x -> ADAM(x, .0005))
+x = RLSetup(learner, env, ConstantNumberEpisodes(10),
+            callbacks = [Progress(), EvaluationPerEpisode(TimeSteps()),
+                         Visualize(wait = 0)])
+info("Before learning.")
+run!(x)
+pop!(x.callbacks)
+x.stoppingcriterion = ConstantNumberEpisodes(400)
+@time learn!(x)
+x.stoppingcriterion = ConstantNumberEpisodes(10)
+push!(x.callbacks, Visualize(wait = 0))
+info("After learning.")
+run!(x)

src/RLEnvGym.jl

@@ -1,7 +1,57 @@
 module RLEnvGym
 using Reexport
 @reexport using ReinforcementLearning
-import ReinforcementLearning:interact!, reset!, getstate
+import ReinforcementLearning:interact!, reset!, getstate, plotenv
+using PyCall
+@pyimport gym
+# @pyimport roboschool
 
+function getspace(space)
+    if pyisinstance(space, gym.spaces[:box][:Box])
+        ReinforcementLearning.Box(space[:low], space[:high])
+    elseif pyisinstance(space, gym.spaces[:discrete][:Discrete])
+        1:space[:n]
+    else
+        error("Don't know how to convert $(pytypeof(space)).")
+    end
+end
+mutable struct GymEnvState
+    done::Bool
+end
+struct GymEnv{TObject, TObsSpace, TActionSpace}
+    pyobj::TObject
+    observation_space::TObsSpace
+    action_space::TActionSpace
+    state::GymEnvState
+end
+function GymEnv(name::String)
+    pyenv = gym.make(name)
+    obsspace = getspace(pyenv[:observation_space])
+    actspace = getspace(pyenv[:action_space])
+    env = GymEnv(pyenv, obsspace, actspace, GymEnvState(false))
+    reset!(env)
+    env
+end
+
+function interactgym!(action, env)
+    if env.state.done 
+        s = reset!(env)
+        r = 0
+        d = false
+    else
+        s, r, d = env.pyobj[:step](action)
+    end
+    env.state.done = d
+    s, r, d
+end
+interact!(action, env::GymEnv) = interactgym!(action, env)
+interact!(action::Int64, env::GymEnv) = interactgym!(action - 1, env)
+reset!(env::GymEnv) = env.pyobj[:reset]()
+getstate(env::GymEnv) = (env.pyobj[:env][:state], false) # doesn't work for all envs
+
+plotenv(env::GymEnv, s, a, r, d) = env.pyobj[:render]()
+listallenvs() = gym.envs[:registry][:all]()
+
+export GymEnv, listallenvs
 
 end # module