Updates to fix issues associated with Flux 0.14 (#76)

* updates to fix issues associated with Flux 0.14

* version bump

.gitignore

@@ -6,4 +6,5 @@ log*
 *.bson
 events.out.tfevents*
 .vscode
-Manifest.toml
+Manifest.toml
+.DS_Store

Project.toml

@@ -1,7 +1,7 @@
 name = "DeepQLearning"
 uuid = "de0a67f4-c691-11e8-0034-5fc6e16e22d3"
 repo = "https://github.com/JuliaPOMDP/DeepQLearning.jl"
-version = "0.7.1"
+version = "0.7.2"
 
 [deps]
 BSON = "fbb218c0-5317-5bc6-957e-2ee96dd4b1f0"

README.md

@@ -27,7 +27,6 @@ using DeepQLearning
 using POMDPs
 using Flux
 using POMDPModels
-using POMDPSimulators
 using POMDPTools
 
 # load MDP model from POMDPModels or define your own!
@@ -37,7 +36,7 @@ mdp = SimpleGridWorld();
 # the gridworld state is represented by a 2 dimensional vector.
 model = Chain(Dense(2, 32), Dense(32, length(actions(mdp))))
 
-exploration = EpsGreedyPolicy(mdp, LinearDecaySchedule(start=1.0, stop=0.01, steps=10000/2))
+exploration = EpsGreedyPolicy(mdp, LinearDecaySchedule(start=1.0, stop=0.01, steps=10000/2));
 
 solver = DeepQLearningSolver(qnetwork = model, max_steps=10000, 
                              exploration_policy = exploration,
@@ -99,39 +98,40 @@ mdp = SimpleGridWorld();
 # the model weights will be send to the gpu in the call to solve
 model = Chain(Dense(2, 32), Dense(32, length(actions(mdp))))
 
-solver = DeepQLearningSolver(qnetwork = model, max_steps=10000, 
-                             learning_rate=0.005,log_freq=500,
-                             recurrence=false,double_q=true, dueling=true, prioritized_replay=true)
+exploration = EpsGreedyPolicy(mdp, LinearDecaySchedule(start=1.0, stop=0.01, steps=10000/2));
+
+solver = DeepQLearningSolver(qnetwork=model, max_steps=10000, 
+                            exploration_policy=exploration,
+                            learning_rate=0.005,log_freq=500,
+                            recurrence=false,double_q=true, dueling=true, prioritized_replay=true)
 policy = solve(solver, mdp)
 ```
 
 ## Solver Options
 
 **Fields of the Q Learning solver:**
 - `qnetwork::Any = nothing` Specify the architecture of the Q network 
+- `exploration_policy::<ExplorationPolicy` Exploration strategy (e.g. EpsGreedyPolicy)
 - `learning_rate::Float64 = 1e-4` learning rate 
 - `max_steps::Int64` total number of training step default = 1000
-- `target_update_freq::Int64` frequency at which the target network is updated default = 500
 - `batch_size::Int64` batch size sampled from the replay buffer default = 32
 - `train_freq::Int64` frequency at which the active network is updated default  = 4
-- `log_freq::Int64` frequency at which to logg info default = 100
 - `eval_freq::Int64` frequency at which to eval the network default = 100
+- `target_update_freq::Int64` frequency at which the target network is updated default = 500
 - `num_ep_eval::Int64` number of episodes to evaluate the policy default = 100
-- `eps_fraction::Float64` fraction of the training set used to explore default = 0.5
-- `eps_end::Float64` value of epsilon at the end of the exploration phase default = 0.01
 - `double_q::Bool` double q learning udpate default = true
 - `dueling::Bool` dueling structure for the q network default = true
 - `recurrence::Bool = false` set to true to use DRQN, it will throw an error if you set it to false and pass a recurrent model.
+- `evaluation_policy::Function = basic_evaluation` function use to evaluate the policy every `eval_freq` steps, the default is a rollout that return the undiscounted average reward 
 - `prioritized_replay::Bool` enable prioritized experience replay default = true
 - `prioritized_replay_alpha::Float64` default = 0.6
 - `prioritized_replay_epsilon::Float64` default = 1e-6
 - `prioritized_replay_beta::Float64` default = 0.4
 - `buffer_size::Int64` size of the experience replay buffer default = 1000
 - `max_episode_length::Int64` maximum length of a training episode default = 100
 - `train_start::Int64` number of steps used to fill in the replay buffer initially default = 200
-- `save_freq::Int64` save the model every `save_freq` steps, default = 1000
-- `evaluation_policy::Function = basic_evaluation` function use to evaluate the policy every `eval_freq` steps, the default is a rollout that return the undiscounted average reward 
-- `exploration_policy::Any = linear_epsilon_greedy(max_steps, eps_fraction, eps_end)` exploration strategy (default is epsilon greedy with linear decay)
 - `rng::AbstractRNG` random number generator default = MersenneTwister(0)
 - `logdir::String = ""` folder in which to save the model
+- `save_freq::Int64` save the model every `save_freq` steps, default = 1000
+- `log_freq::Int64` frequency at which to logg info default = 100
 - `verbose::Bool` default = true

src/dueling.jl

@@ -10,7 +10,7 @@ function (m::DuelingNetwork)(inpt)
     return m.val(x) .+ m.adv(x) .- mean(m.adv(x), dims=1)
 end
 
-Flux.@functor DuelingNetwork
+Flux.@layer DuelingNetwork
 
 function Flux.reset!(m::DuelingNetwork)
     Flux.reset!(m.base)

src/solver.jl

@@ -1,5 +1,6 @@
 @with_kw mutable struct DeepQLearningSolver{E<:ExplorationPolicy} <: Solver
     qnetwork::Any = nothing # intended to be a flux model
+    exploration_policy::E # No default since 9ac3ab
     learning_rate::Float32 = 1f-4
     max_steps::Int64 = 1000
     batch_size::Int64 = 32
@@ -11,7 +12,6 @@
     dueling::Bool = true
     recurrence::Bool = false
     evaluation_policy::Any = basic_evaluation
-    exploration_policy::E
     trace_length::Int64 = 40
     prioritized_replay::Bool = true
     prioritized_replay_alpha::Float32 = 0.6f0
@@ -139,9 +139,8 @@ function dqn_train!(solver::DeepQLearningSolver, env::AbstractEnv, policy::Abstr
             sethiddenstates!(active_q, hs)
         end
 
-        if t%solver.target_update_freq == 0
-            weights = Flux.params(active_q)
-            Flux.loadparams!(target_q, weights)
+        if t % solver.target_update_freq == 0
+            target_q = deepcopy(active_q)
         end
 
         if t % solver.eval_freq == 0
@@ -170,9 +169,9 @@ function dqn_train!(solver::DeepQLearningSolver, env::AbstractEnv, policy::Abstr
     if model_saved
         if solver.verbose
             @printf("Restore model with eval reward %1.3f \n", saved_mean_reward)
-            saved_model = BSON.load(joinpath(solver.logdir, "qnetwork.bson"))[:qnetwork]
-            Flux.loadparams!(getnetwork(policy), saved_model)
         end
+        saved_model_state = BSON.load(joinpath(solver.logdir, "qnetwork_state.bson"))[:qnetwork_state]
+        Flux.loadmodel!(policy.qnetwork, saved_model_state)
     end
     return policy
 end
@@ -289,7 +288,9 @@ end
 
 function save_model(solver::DeepQLearningSolver, active_q, scores_eval::Float64, saved_mean_reward::Float64, model_saved::Bool)
     if scores_eval >= saved_mean_reward
-        bson(joinpath(solver.logdir, "qnetwork.bson"), qnetwork=[w for w in Flux.params(active_q)])
+        copied_model = deepcopy(active_q)
+        Flux.reset!(copied_model)
+        bson(joinpath(solver.logdir, "qnetwork_state.bson"), qnetwork_state=Flux.state(copied_model))
         if solver.verbose
             @printf("Saving new model with eval reward %1.3f \n", scores_eval)
         end
@@ -311,8 +312,8 @@ function restore_best_model(solver::DeepQLearningSolver, env::AbstractEnv)
         active_q = solver.qnetwork
     end
     policy = NNPolicy(env, active_q, collect(actions(env)), length(obs_dimensions(env)))
-    weights = BSON.load(solver.logdir*"qnetwork.bson")[:qnetwork]
-    Flux.loadparams!(getnetwork(policy), weights)
+    saved_network_state = BSON.load(solver.logdir*"qnetwork_state.bson")[:qnetwork_state]
+    Flux.loadmodel!(getnetwork(policy), saved_network_state)
     Flux.testmode!(getnetwork(policy))
     return policy
 end

test/README_examples.jl

@@ -0,0 +1,42 @@
+using DeepQLearning
+using POMDPs
+using Flux
+using POMDPModels
+using POMDPTools
+
+@testset "README Example 1" begin
+    # load MDP model from POMDPModels or define your own!
+    mdp = SimpleGridWorld();
+
+    # Define the Q network (see Flux.jl documentation)
+    # the gridworld state is represented by a 2 dimensional vector.
+    model = Chain(Dense(2, 32), Dense(32, length(actions(mdp))))
+
+    exploration = EpsGreedyPolicy(mdp, LinearDecaySchedule(start=1.0, stop=0.01, steps=10000/2));
+
+    solver = DeepQLearningSolver(qnetwork = model, max_steps=10000, 
+                                exploration_policy = exploration,
+                                learning_rate=0.005,log_freq=500,
+                                recurrence=false,double_q=true, dueling=true, prioritized_replay=true)
+    policy = solve(solver, mdp)
+
+    sim = RolloutSimulator(max_steps=30)
+    r_tot = simulate(sim, mdp, policy)
+    println("Total discounted reward for 1 simulation: $r_tot")
+end
+
+@testset "README Example 2" begin
+    # Without using CuArrays
+    mdp = SimpleGridWorld();
+
+    # the model weights will be send to the gpu in the call to solve
+    model = Chain(Dense(2, 32), Dense(32, length(actions(mdp))))
+
+    exploration = EpsGreedyPolicy(mdp, LinearDecaySchedule(start=1.0, stop=0.01, steps=10000/2));
+
+    solver = DeepQLearningSolver(qnetwork=model, max_steps=10000, 
+                                exploration_policy=exploration,
+                                learning_rate=0.005,log_freq=500,
+                                recurrence=false,double_q=true, dueling=true, prioritized_replay=true)
+    policy = solve(solver, mdp)
+end

test/runtests.jl

@@ -232,3 +232,8 @@ end
 
     @test evaluate(env, policy, GLOBAL_RNG) > 1.0
 end
+
+
+@testset "README Examples" begin
+    include("README_examples.jl")
+end