manager.initialize_aggregator(path=saving_path,
saving_after=5,
aggregator_keys=["loss", 'reward', 'time'])
rewards = []
# Get initial agent
agent = manager.get_agent()
print('TRAINING')
for e in range(max_episodes):
# Sample data to optimize
print('sampling...')
sample_dict = manager.sample(sample_size=sampled_batches *
optimization_batch_size,
from_buffer=False)
# Compute Advantages
print('calculate advantage estimates...')
# Add value of last 'new_state'
sample_dict['value_estimate'].append(
agent.v_estimate(np.expand_dims(sample_dict['state_new'][-1], 0)))
sample_dict['advantage'] = []
gae = 0
# Loop backwards through rewards
for i in reversed(range(len(sample_dict['reward']))):
delta = sample_dict['reward'][i] + gamma * sample_dict[
'value_estimate'][i + 1].numpy() * sample_dict['not_done'][
rewards = []
# Get initial agent
agent = manager.get_agent()
print('TRAINING')
for e in range(max_episodes):
# Experience replay
print("collecting experience..")
print('Temperature: ',temperature)
data = manager.get_data()
manager.store_in_buffer(data)
# Sample data to optimize on from buffer
sample_dict = manager.sample(sampled_batches*optimization_batch_size)
samples,_ = dict_to_dataset(sample_dict,batch_size = optimization_batch_size)
print("optimizing...")
losses = []
for s, a, r, s_1, not_done in samples:
a = tf.cast(a, dtype = tf.int32)
target = r + gamma*agent.max_q(s_1)*not_done
with tf.GradientTape() as tape:
pred = agent.q_val(s,a)
loss = loss_function(target,pred)
gradients = tape.gradient(loss, agent.model.trainable_variables)
optimizer.apply_gradients(zip(gradients, agent.model.trainable_variables))
losses.append(loss)
manager.test(test_steps, test_episodes=10, do_print=True, render=True)
# get initial agent
agent = manager.get_agent()
for e in range(epochs):
# training core
# experience replay
print("collecting experience..")
data = manager.get_data(total_steps=100)
manager.store_in_buffer(data)
# sample data to optimize on from buffer
sample_dict = manager.sample(sample_size)
print(f"collected data for: {sample_dict.keys()}")
# create and batch tf datasets
data_dict = dict_to_dict_of_datasets(sample_dict, batch_size=64)
print("optimizing...")
# for each batch
for state, action, reward, state_new, not_done in zip(
data_dict['state'], data_dict['action'], data_dict['reward'],
data_dict['state_new'], data_dict['not_done']):
q_target = tf.cast(reward, tf.float64) + (
tf.cast(not_done, tf.float64) *
tf.cast(gamma * agent.max_q(state_new), tf.float64))
with tf.GradientTape() as tape:
prediction = agent.q_val(state, action)
# initial testing:
print('test before training: ')
manager.test(test_steps,
test_episodes=10,
do_print=True,
evaluation_measure='time_and_reward')
for e in range(epochs):
print('collecting experience..')
data = manager.get_data()
manager.store_in_buffer(data)
# sample data to optimize on from buffer
experience_dict = manager.sample(sample_size)
print('optimizing...')
for states, actions, rewards, states_new, not_dones in zip(
*[experience_dict[k] for k in optim_keys]):
train_step(agent.model, states, actions, rewards, states_new,
not_dones, learning_rate, gamma)
# set new weights, get optimized agent
manager.set_agent(agent.model.get_weights())
# update aggregator
time_steps, reward_agg = manager.test(
test_steps, evaluation_measure='time_and_reward')
manager.update_aggregator(time_steps=time_steps, rewards=reward_agg)
test_steps = 1000
# Factor of how much the new policy is allowed to differ from the old one
epsilon = 0.2
entropy_weight = 0.01
# initilize progress aggregator
manager.initialize_aggregator(path=saving_path,
saving_after=5,
aggregator_keys=["loss", "reward"])
agent = manager.get_agent()
optimizer = tf.keras.optimizers.Adam(learning_rate=.0001)
for e in range(epochs):
sample_dict = manager.sample(sample_size, from_buffer=False)
print(f"collected data for: {sample_dict.keys()}")
# Shift value estimate by one to the left to get the value estimate of next state
state_value = tf.squeeze(sample_dict['value_estimate'])
state_value_new = tf.roll(state_value, -1, axis=0)
not_done = tf.cast(sample_dict['not_done'], tf.bool)
state_value_new = tf.where(not_done, state_value_new, 0)
# Calculate advantate estimate q(s,a)-b(s)=r+v(s')-v(s)
advantage_estimate = -state_value + sample_dict[
'reward'] + gamma * state_value_new
sample_dict['advantage_estimate'] = advantage_estimate
data_dict = dict_to_dict_of_datasets(sample_dict,
batch_size=optim_batch_size)
def train_td3(args, model, action_dimension=None):
print(args)
tf.keras.backend.set_floatx('float32')
ray.init(log_to_driver=False)
# hyper parameters
buffer_size = args.buffer_size # 10e6 in their repo, not possible with our ram
epochs = args.epochs
saving_path = os.getcwd() + "/" + args.saving_dir
saving_after = 5
sample_size = args.sample_size
optim_batch_size = args.batch_size
gamma = args.gamma
test_steps = 100 # 1000 in their repo
policy_delay = 2
rho = .046
policy_noise = args.policy_noise
policy_noise_clip = .5
msg_dim = args.msg_dim # 32 in their repo
learning_rate = args.learning_rate
save_args(args, saving_path)
env_test_instance = gym.make('BipedalWalker-v3')
if action_dimension is None:
action_dimension = copy(env_test_instance.action_space.shape[0])
model_kwargs = {
# action dimension for modular actions
'action_dimension': action_dimension,
'min_action': copy(env_test_instance.action_space.low)[0],
'max_action': copy(env_test_instance.action_space.high)[0],
'msg_dimension': msg_dim,
'fix_sigma': True,
'hidden_units': args.hidden_units
}
del env_test_instance
manager = SampleManager(model,
'BipedalWalker-v3',
num_parallel=(os.cpu_count() - 1),
total_steps=150,
action_sampling_type="continuous_normal_diagonal",
is_tf=True,
model_kwargs=model_kwargs)
optim_keys = [
'state',
'action',
'reward',
'state_new',
'not_done',
]
manager.initialize_buffer(buffer_size, optim_keys)
manager.initialize_aggregator(path=saving_path,
saving_after=saving_after,
aggregator_keys=["loss", "reward"])
agent = manager.get_agent()
optimizer = tf.keras.optimizers.Adam(learning_rate=learning_rate)
# fill buffer
print("Filling buffer before training..")
while len(manager.buffer.buffer[
manager.buffer.keys[0]]) < manager.buffer.size:
# Gives you state action reward trajectories
data = manager.get_data()
manager.store_in_buffer(data)
# track time while training
timer = time.time()
last_t = timer
target_agent = manager.get_agent()
for e in range(epochs):
# off policy
sample_dict = manager.sample(sample_size, from_buffer=True)
print(f"collected data for: {sample_dict.keys()}")
# cast values to float32 and create data dict
sample_dict['state'] = tf.cast(sample_dict['state'], tf.float32)
sample_dict['action'] = tf.cast(sample_dict['action'], tf.float32)
sample_dict['reward'] = tf.cast(sample_dict['reward'], tf.float32)
sample_dict['state_new'] = tf.cast(sample_dict['state_new'],
tf.float32)
sample_dict['not_done'] = tf.cast(sample_dict['not_done'], tf.float32)
data_dict = dict_to_dict_of_datasets(sample_dict,
batch_size=optim_batch_size)
total_loss = 0
for state, action, reward, state_new, not_done in \
zip(data_dict['state'],
data_dict['action'],
data_dict['reward'],
data_dict['state_new'],
data_dict['not_done']):
action_new = target_agent.act(state_new)
# add noise to action_new
action_new = action_new + tf.clip_by_value(
tf.random.normal(action_new.shape, 0., policy_noise),
-policy_noise_clip, policy_noise_clip)
# clip action_new to action space
action_new = tf.clip_by_value(
action_new, manager.env_instance.action_space.low,
manager.env_instance.action_space.high)
# calculate target with double-Q-learning
state_action_new = tf.concat([state_new, action_new], axis=-1)
q_values0 = target_agent.model.critic0(state_action_new)
q_values1 = target_agent.model.critic1(state_action_new)
q_values = tf.concat([q_values0, q_values1], axis=-1)
q_targets = tf.squeeze(tf.reduce_min(q_values, axis=-1))
critic_target = reward + gamma * not_done * q_targets
state_action = tf.concat([state, action], axis=-1)
# update critic 0
with tf.GradientTape() as tape:
q_output = agent.model.critic0(state_action)
loss = tf.keras.losses.MSE(tf.squeeze(critic_target),
tf.squeeze(q_output))
total_loss += loss
gradients = tape.gradient(loss,
agent.model.critic0.trainable_variables)
optimizer.apply_gradients(
zip(gradients, agent.model.critic0.trainable_variables))
# update critic 1
with tf.GradientTape() as tape:
q_output = agent.model.critic1(state_action)
loss = tf.keras.losses.MSE(tf.squeeze(critic_target),
tf.squeeze(q_output))
total_loss += loss
gradients = tape.gradient(loss,
agent.model.critic1.trainable_variables)
optimizer.apply_gradients(
zip(gradients, agent.model.critic1.trainable_variables))
# update actor with delayed policy update
if e % policy_delay == 0:
with tf.GradientTape() as tape:
actor_output = agent.model.actor(state)
action = reparam_action(actor_output,
agent.model.action_dimension,
agent.model.min_action,
agent.model.max_action)
state_action = tf.concat([state, action], axis=-1)
q_val = agent.model.critic0(state_action)
actor_loss = -tf.reduce_mean(q_val)
total_loss += actor_loss
actor_gradients = tape.gradient(
actor_loss, agent.model.actor.trainable_variables)
optimizer.apply_gradients(
zip(actor_gradients,
agent.model.actor.trainable_variables))
# Update agent
manager.set_agent(agent.get_weights())
agent = manager.get_agent()
if e % policy_delay == 0:
# Polyak averaging
new_weights = list(rho * np.array(target_agent.get_weights()) +
(1. - rho) * np.array(agent.get_weights()))
target_agent.set_weights(new_weights)
reward = manager.test(test_steps, evaluation_measure="reward")
manager.update_aggregator(loss=total_loss, reward=reward)
print(
f"epoch ::: {e} loss ::: {total_loss} avg reward ::: {np.mean(reward)}"
)
if e % saving_after == 0:
manager.save_model(saving_path, e)
# needed time and remaining time estimation
current_t = time.time()
time_needed = (current_t - last_t) / 60.
time_remaining = (current_t - timer) / 60. / (e + 1) * (epochs -
(e + 1))
print(
'Finished epoch %d of %d. Needed %1.f min for this epoch. Estimated time remaining: %.1f min'
% (e + 1, epochs, time_needed, time_remaining))
last_t = current_t
manager.load_model(saving_path)
print("done")
print("testing optimized agent")
manager.test(test_steps, test_episodes=10, render=True)
ray.shutdown()
print('# =============== INITIAL TESTING =============== #')
manager.test(MAX_TEST_STEPS, 5, evaluation_measure='time_and_reward', do_print=True, render=True)
# get the initial agent
agent = manager.get_agent()
print('# =============== START TRAINING ================ #')
for e in range(1, EPOCHS+1):
print(f'# ============== EPOCH {e}/{EPOCHS} ============== #')
print('# ============= collecting samples ============== #')
# collect experience and save it to ERP buffer
data = manager.get_data(do_print=False)
manager.store_in_buffer(data)
# get some samples from the ERP buffer and create a dataset
sample_dict = manager.sample(sample_size=SAMPLE_SIZE)
data_dict = dict_to_dict_of_datasets(sample_dict, batch_size=BATCH_SIZE)
dataset = tf.data.Dataset.zip((data_dict['state'], data_dict['action'], data_dict['reward'], data_dict['state_new'], data_dict['not_done']))
print('# ================= optimizing ================== #')
losses = []
for s, a, r, ns, nd in dataset:
# ensure that the datasets have at least 10 elements
# otherwise we run into problems with the MSE loss
if len(s) >= 10:
loss = train_q_network(agent, s, a, r, ns, nd, optimizer)
losses.append(loss)
print(f'average loss: {np.mean(losses)}')
print("Training the agent.")
# get initial agent
agent = manager.get_agent()
# circular buffer for avg env steps (used to stop training if agent is good)
mean_list = collections.deque(maxlen=3)
for e in range(epochs):
data = manager.get_data(total_steps=1000)
manager.store_in_buffer(data)
# sample data to optimize on from buffer
sample_dict = manager.sample(sample_size, from_buffer=True)
# create and batch tf datasets
data_dict = dict_to_dict_of_datasets(sample_dict,
batch_size=optim_batch_size)
loss = 0.0
for state, action, reward, state_next, nd in zip(
data_dict['state'], data_dict['action'], data_dict['reward'],
data_dict['state_new'], data_dict['not_done']):
q_target = tf.cast(reward, tf.float64) + (
tf.cast(nd, tf.float64) *
tf.cast(gamma * agent.max_q(state_next), tf.float64))
# apply backpropagation and sum up the losses