# --- Instantiate qnetwork ---
qnetwork = MyQNetwork(
environment=env,
random_state=rng)
# --- Instantiate agent ---
agent = NeuralAgent(
env,
qnetwork,
random_state=rng)
# --- Bind controllers to the agent ---
# Before every training epoch, we want to print a summary of the agent's epsilon, discount and
# learning rate as well as the training epoch number.
agent.attach(bc.VerboseController())
# During training epochs, we want to train the agent after every action it takes.
# Plus, we also want to display after each training episode (!= than after every training) the average bellman
# residual and the average of the V values obtained during the last episode.
agent.attach(bc.TrainerController())
# All previous controllers control the agent during the epochs it goes through. However, we want to interleave a
# "test epoch" between each training epoch. We do not want these test epoch to interfere with the training of the
# agent. Therefore, we will disable these controllers for the whole duration of the test epochs interleaved this
# way, using the controllersToDisable argument of the InterleavedTestEpochController. The value of this argument
# is a list of the indexes of all controllers to disable, their index reflecting in which order they were added.
agent.attach(bc.InterleavedTestEpochController(
epoch_length=500,
controllers_to_disable=[0, 1]))
double_Q=True)
# --- Instantiate agent ---
agent = NeuralAgent(
env,
qnetwork,
parameters.replay_memory_size,
max(env.inputDimensions()[i][0] for i in range(len(env.inputDimensions()))),
parameters.batch_size,
rng,
exp_priority=1.)
# --- Bind controllers to the agent ---
# For comments, please refer to run_toy_env.py
agent.attach(bc.VerboseController(
evaluate_on='epoch',
periodicity=1))
agent.attach(bc.TrainerController(
evaluate_on='action',
periodicity=parameters.update_frequency,
show_episode_avg_V_value=True,
show_avg_Bellman_residual=True))
agent.attach(bc.LearningRateController(
initial_learning_rate=parameters.learning_rate,
learning_rate_decay=parameters.learning_rate_decay,
periodicity=1))
agent.attach(bc.DiscountFactorController(
initial_discount_factor=parameters.discount,
rng,
DoubleQ=True)
# --- Instantiate agent ---
agent = NeuralAgent(
env,
qnetwork,
parameters.replay_memory_size,
max(env.inputDimensions()[i][0] for i in range(len(env.inputDimensions()))),
parameters.batch_size,
rng)
# --- Bind controllers to the agent ---
# For comments, please refer to run_toy_env.py
agent.attach(bc.VerboseController(
evaluateOn='epoch',
periodicity=1))
agent.attach(bc.TrainerController(
evaluateOn='action',
periodicity=parameters.update_frequency,
showEpisodeAvgVValue=False,
showAvgBellmanResidual=False))
agent.attach(bc.LearningRateController(
initialLearningRate=parameters.learning_rate,
learningRateDecay=parameters.learning_rate_decay,
periodicity=1))
agent.attach(bc.DiscountFactorController(
initialDiscountFactor=parameters.discount,
max(env.inputDimensions()[i][0] for i in range(len(env.inputDimensions()))),
parameters.batch_size,
rng,
test_policy=test_policy,
)
# --- Create unique filename for FindBestController ---
h = hash(vars(parameters), hash_name="sha1")
fname = "ALE_" + h
print("The parameters hash is: {}".format(h))
print("The parameters are: {}".format(parameters))
# --- Bind controllers to the agent ---
# Before every training epoch (periodicity=1), we want to print a summary of the agent's epsilon, discount and
# learning rate as well as the training epoch number.
agent.attach(bc.VerboseController(evaluate_on="epoch", periodicity=1))
# During training epochs, we want to train the agent after every [parameters.update_frequency] action it takes.
# Plus, we also want to display after each training episode (!= than after every training) the average bellman
# residual and the average of the V values obtained during the last episode, hence the two last arguments.
agent.attach(
bc.TrainerController(
evaluate_on="action",
periodicity=parameters.update_frequency,
show_episode_avg_V_value=True,
show_avg_Bellman_residual=True,
)
)
# Every epoch end, one has the possibility to modify the learning rate using a LearningRateController. Here we
# wish to update the learning rate after every training epoch (periodicity=1), according to the parameters given.
max(env.inputDimensions()[i][0]
for i in range(len(env.inputDimensions()))),
parameters.batch_size,
rng,
test_policy=test_policy)
# --- Create unique filename for FindBestController ---
h = hash(vars(parameters), hash_name="sha1")
fname = "test_" + h
print("The parameters hash is: {}".format(h))
print("The parameters are: {}".format(parameters))
# --- Bind controllers to the agent ---
# Before every training epoch (periodicity=1), we want to print a summary of the agent's epsilon, discount and
# learning rate as well as the training epoch number.
agent.attach(bc.VerboseController(evaluate_on='epoch', periodicity=1))
# As for the discount factor and the learning rate, one can update periodically the parameter of the epsilon-greedy
# policy implemented by the agent. This controllers has a bit more capabilities, as it allows one to choose more
# precisely when to update epsilon: after every X action, episode or epoch. This parameter can also be reset every
# episode or epoch (or never, hence the resetEvery='none').
agent.attach(
bc.EpsilonController(initial_e=parameters.epsilon_start,
e_decays=parameters.epsilon_decay,
e_min=parameters.epsilon_min,
evaluate_on='action',
periodicity=1,
reset_every='none'))
# --- Run the experiment ---
try:
# --- Instantiate agent ---
agent = NeuralAgent(env,
qnetwork,
parameters.replay_memory_size,
max(env.inputDimensions()[i][0]
for i in range(len(env.inputDimensions()))),
parameters.batch_size,
rng,
exp_priority=1.,
train_policy=train_policy,
test_policy=test_policy)
# --- Bind controllers to the agent ---
# For comments, please refer to run_toy_env.py
agent.attach(bc.VerboseController(evaluate_on='epoch', periodicity=1))
agent.attach(
bc.TrainerController(evaluate_on='action',
periodicity=parameters.update_frequency,
show_episode_avg_V_value=True,
show_avg_Bellman_residual=True))
agent.attach(
bc.LearningRateController(
initial_learning_rate=parameters.learning_rate,
learning_rate_decay=parameters.learning_rate_decay,
periodicity=1))
agent.attach(
bc.DiscountFactorController(
rng = np.random.RandomState(123456)
# --- Instantiate environment ---
env = Toy_env(rng)
# --- Instantiate qnetwork ---
qnetwork = MyQNetwork(environment=env, random_state=rng)
# --- Instantiate agent ---
agent = NeuralAgent(env, qnetwork, random_state=rng)
# --- Bind controllers to the agent ---
# Before every training epoch, we want to print a summary of the agent's epsilon, discount and
# learning rate as well as the training epoch number.
agent.attach(bc.VerboseController())
# During training epochs, we want to train the agent after every action it takes.
# Plus, we also want to display after each training episode (!= than after every training) the average bellman
# residual and the average of the V values obtained during the last episode.
agent.attach(bc.TrainerController())
# All previous controllers control the agent during the epochs it goes through. However, we want to interleave a
# "test epoch" between each training epoch. We do not want these test epoch to interfere with the training of the
# agent. Therefore, we will disable these controllers for the whole duration of the test epochs interleaved this
# way, using the controllersToDisable argument of the InterleavedTestEpochController. The value of this argument
# is a list of the indexes of all controllers to disable, their index reflecting in which order they were added.
agent.attach(
bc.InterleavedTestEpochController(epoch_length=500,
controllers_to_disable=[0, 1]))
rng = np.random.RandomState(123456) # --- Instantiate environment --- env = Toy_env(rng) # --- Instantiate qnetwork --- qnetwork = MyQNetwork(environment=env, random_state=rng) # --- Instantiate agent --- agent = NeuralAgent(env, qnetwork, random_state=rng) # --- Bind controllers to the agent --- # Before every training epoch, we want to print a summary of the agent's epsilon, discount and # learning rate as well as the training epoch number. agent.attach(bc.VerboseController()) # During training epochs, we want to train the agent after every action it takes. # Plus, we also want to display after each training episode (!= than after every training) the average bellman # residual and the average of the V values obtained during the last episode. agent.attach(bc.TrainerController()) # All previous controllers control the agent during the epochs it goes through. However, we want to interleave a # "test epoch" between each training epoch. We do not want these test epoch to interfere with the training of the # agent. Therefore, we will disable these controllers for the whole duration of the test epochs interleaved this # way, using the controllersToDisable argument of the InterleavedTestEpochController. The value of this argument # is a list of the indexes of all controllers to disable, their index reflecting in which order they were added. agent.attach(bc.InterleavedTestEpochController(epoch_length=500)) # --- Run the experiment --- agent.run(n_epochs=100, epoch_length=1000)
rng,
DoubleQ=True)
# --- Instantiate agent ---
agent = NeuralAgent(
env,
qnetwork,
parameters.replay_memory_size,
max(env.inputDimensions()[i][0] for i in range(len(env.inputDimensions()))),
parameters.batch_size,
rng)
# --- Bind controllers to the agent ---
# For comments, please refer to run_toy_env.py
agent.attach(bc.VerboseController(
evaluateOn='epoch',
periodicity=1))
agent.attach(bc.TrainerController(
evaluateOn='action',
periodicity=parameters.update_frequency,
showEpisodeAvgVValue=True,
showAvgBellmanResidual=True))
agent.attach(bc.LearningRateController(
initialLearningRate=parameters.learning_rate,
learningRateDecay=parameters.learning_rate_decay,
periodicity=1))
agent.attach(bc.DiscountFactorController(
initialDiscountFactor=parameters.discount,
