def train_epoch(self, epoch_info: EpochInfo) -> None:
""" Train model on an epoch of a fixed number of batch updates """
epoch_info.on_epoch_begin()
for batch_idx in tqdm.trange(epoch_info.batches_per_epoch,
file=sys.stdout,
desc="Training",
unit="batch"):
batch_info = BatchInfo(epoch_info, batch_idx)
batch_info.on_batch_begin()
self.train_batch(batch_info)
batch_info.on_batch_end()
epoch_info.result_accumulator.freeze_results()
epoch_info.on_epoch_end()
def train_model():
"""a sample training script, that creates a PPO instance and train it with bc-gym environment
:return: None
"""
device = torch.device('cpu')
seed = 1001
# Set random seed in python std lib, numpy and pytorch
set_seed(seed)
env_function = lambda: ColoredEgoCostmapRandomAisleTurnEnv()
vec_env = DummyVecEnv([env_function])
# Again, use a helper to create a model
# But because model is owned by the reinforcer, model should not be accessed using this variable
# but from reinforcer.model property
model = PolicyGradientModelFactory(backbone=NatureCnnTwoTowerFactory(
input_width=133, input_height=133, input_channels=1)).instantiate(
action_space=vec_env.action_space)
# Set schedule for gradient clipping.
cliprange = LinearSchedule(initial_value=0.01, final_value=0.0)
# Reinforcer - an object managing the learning process
reinforcer = OnPolicyIterationReinforcer(
device=device,
settings=OnPolicyIterationReinforcerSettings(discount_factor=0.99,
batch_size=256,
experience_replay=4),
model=model,
algo=PpoPolicyGradient(entropy_coefficient=0.01,
value_coefficient=0.5,
max_grad_norm=0.01,
cliprange=cliprange),
env_roller=StepEnvRoller(
environment=vec_env,
device=device,
gae_lambda=0.95,
number_of_steps=128,
discount_factor=0.99,
))
# Model optimizer
optimizer = optim.Adam(reinforcer.model.parameters(), lr=1e-6, eps=1.0e-5)
# Overall information store for training information
training_info = TrainingInfo(
metrics=[
EpisodeRewardMetric(
'episode_rewards'), # Calculate average reward from episode
],
callbacks=[
StdoutStreaming(
), # Print live metrics every epoch to standard output
FrameTracker(
1.1e8
) # We need frame tracker to track the progress of learning
])
# A bit of training initialization bookkeeping...
training_info.initialize()
reinforcer.initialize_training(training_info)
training_info.on_train_begin()
# Let's make 10 batches per epoch to average metrics nicely
# Rollout size is 8 environments times 128 steps
num_epochs = int(1.1e8 / (128 * 1) / 10)
# Normal handrolled training loop
eval_results = []
for i in range(1, num_epochs + 1):
epoch_info = EpochInfo(training_info=training_info,
global_epoch_idx=i,
batches_per_epoch=10,
optimizer=optimizer)
reinforcer.train_epoch(epoch_info)
eval_result = evaluate_model(model, vec_env, device, takes=1)
eval_results.append(eval_result)
if i % 100 == 0:
torch.save(model.state_dict(), 'tmp_checkout.data')
with open('tmp_eval_results.pkl', 'wb') as f:
pickle.dump(eval_results, f, 0)
training_info.on_train_end()
def breakout_a2c():
device = torch.device('cuda:0')
seed = 1001
# Set random seed in python std lib, numpy and pytorch
set_seed(seed)
# Create 16 environments evaluated in parallel in sub processess with all usual DeepMind wrappers
# These are just helper functions for that
vec_env = SubprocVecEnvWrapper(ClassicAtariEnv('BreakoutNoFrameskip-v4'),
frame_history=4).instantiate(
parallel_envs=16, seed=seed)
# Again, use a helper to create a model
# But because model is owned by the reinforcer, model should not be accessed using this variable
# but from reinforcer.model property
model = PolicyGradientModelFactory(backbone=NatureCnnFactory(
input_width=84, input_height=84, input_channels=4)).instantiate(
action_space=vec_env.action_space)
# Reinforcer - an object managing the learning process
reinforcer = OnPolicyIterationReinforcer(
device=device,
settings=OnPolicyIterationReinforcerSettings(
discount_factor=0.99,
batch_size=256,
),
model=model,
algo=A2CPolicyGradient(entropy_coefficient=0.01,
value_coefficient=0.5,
max_grad_norm=0.5),
env_roller=StepEnvRoller(
environment=vec_env,
device=device,
number_of_steps=5,
discount_factor=0.99,
))
# Model optimizer
optimizer = optim.RMSprop(reinforcer.model.parameters(),
lr=7.0e-4,
eps=1e-3)
# Overall information store for training information
training_info = TrainingInfo(
metrics=[
EpisodeRewardMetric(
'episode_rewards'), # Calculate average reward from episode
],
callbacks=[StdoutStreaming()
] # Print live metrics every epoch to standard output
)
# A bit of training initialization bookkeeping...
training_info.initialize()
reinforcer.initialize_training(training_info)
training_info.on_train_begin()
# Let's make 100 batches per epoch to average metrics nicely
num_epochs = int(1.1e7 / (5 * 16) / 100)
# Normal handrolled training loop
for i in range(1, num_epochs + 1):
epoch_info = EpochInfo(training_info=training_info,
global_epoch_idx=i,
batches_per_epoch=100,
optimizer=optimizer)
reinforcer.train_epoch(epoch_info)
training_info.on_train_end()
def test_acer_breakout():
"""
1 iteration of ACER on breakout environment
"""
device = torch.device('cpu')
seed = 1001
# Set random seed in python std lib, numpy and pytorch
set_seed(seed)
# Create 16 environments evaluated in parallel in sub processess with all usual DeepMind wrappers
# These are just helper functions for that
vec_env = SubprocVecEnvWrapper(ClassicAtariEnv('BreakoutNoFrameskip-v4'),
frame_history=4).instantiate(
parallel_envs=16, seed=seed)
# Again, use a helper to create a model
# But because model is owned by the reinforcer, model should not be accessed using this variable
# but from reinforcer.model property
model_factory = QPolicyGradientModelFactory(backbone=NatureCnnFactory(
input_width=84, input_height=84, input_channels=4))
# Reinforcer - an object managing the learning process
reinforcer = BufferedMixedPolicyIterationReinforcer(
device=device,
settings=BufferedMixedPolicyIterationReinforcerSettings(
discount_factor=0.99,
experience_replay=2,
stochastic_experience_replay=False),
model=model_factory.instantiate(action_space=vec_env.action_space),
env=vec_env,
algo=AcerPolicyGradient(
model_factory=model_factory,
entropy_coefficient=0.01,
q_coefficient=0.5,
rho_cap=10.0,
retrace_rho_cap=1.0,
trust_region=True,
trust_region_delta=1.0,
max_grad_norm=10.0,
),
env_roller=ReplayQEnvRoller(environment=vec_env,
device=device,
number_of_steps=12,
discount_factor=0.99,
buffer_capacity=100,
buffer_initial_size=100,
frame_stack_compensation=4),
)
# Model optimizer
optimizer = optim.RMSprop(reinforcer.model.parameters(),
lr=7.0e-4,
eps=1e-3,
alpha=0.99)
# Overall information store for training information
training_info = TrainingInfo(
metrics=[
EpisodeRewardMetric(
'episode_rewards'), # Calculate average reward from episode
],
callbacks=[] # Print live metrics every epoch to standard output
)
# A bit of training initialization bookkeeping...
training_info.initialize()
reinforcer.initialize_training(training_info)
training_info.on_train_begin()
# Let's make 100 batches per epoch to average metrics nicely
num_epochs = 1
# Normal handrolled training loop
for i in range(1, num_epochs + 1):
epoch_info = EpochInfo(training_info=training_info,
global_epoch_idx=i,
batches_per_epoch=1,
optimizer=optimizer)
reinforcer.train_epoch(epoch_info, interactive=False)
training_info.on_train_end()
def test_trpo_bipedal_walker():
"""
1 iteration of TRPO on bipedal walker
"""
device = torch.device('cpu')
seed = 1001
# Set random seed in python std lib, numpy and pytorch
set_seed(seed)
vec_env = DummyVecEnvWrapper(MujocoEnv('BipedalWalker-v2'),
normalize=True).instantiate(parallel_envs=8,
seed=seed)
# Again, use a helper to create a model
# But because model is owned by the reinforcer, model should not be accessed using this variable
# but from reinforcer.model property
model_factory = PolicyGradientModelSeparateFactory(
policy_backbone=MLPFactory(input_length=24, hidden_layers=[32, 32]),
value_backbone=MLPFactory(input_length=24, hidden_layers=[32]))
# Reinforcer - an object managing the learning process
reinforcer = OnPolicyIterationReinforcer(
device=device,
settings=OnPolicyIterationReinforcerSettings(discount_factor=0.99, ),
model=model_factory.instantiate(action_space=vec_env.action_space),
algo=TrpoPolicyGradient(
max_kl=0.01,
cg_iters=10,
line_search_iters=10,
improvement_acceptance_ratio=0.1,
cg_damping=0.1,
vf_iters=5,
entropy_coef=0.0,
max_grad_norm=0.5,
),
env_roller=StepEnvRoller(
environment=vec_env,
device=device,
number_of_steps=12,
discount_factor=0.99,
))
# Model optimizer
optimizer = optim.Adam(reinforcer.model.parameters(), lr=1.0e-3, eps=1e-4)
# Overall information store for training information
training_info = TrainingInfo(
metrics=[
EpisodeRewardMetric(
'episode_rewards'), # Calculate average reward from episode
],
callbacks=[FrameTracker(100_000)
] # Print live metrics every epoch to standard output
)
# A bit of training initialization bookkeeping...
training_info.initialize()
reinforcer.initialize_training(training_info)
training_info.on_train_begin()
# Let's make 100 batches per epoch to average metrics nicely
num_epochs = 1
# Normal handrolled training loop
for i in range(1, num_epochs + 1):
epoch_info = EpochInfo(training_info=training_info,
global_epoch_idx=i,
batches_per_epoch=1,
optimizer=optimizer)
reinforcer.train_epoch(epoch_info, interactive=False)
training_info.on_train_end()
def test_ddpg_bipedal_walker():
"""
1 iteration of DDPG bipedal walker environment
"""
device = torch.device('cpu')
seed = 1001
# Set random seed in python std lib, numpy and pytorch
set_seed(seed)
# Only single environment for DDPG
env = MujocoEnv('BipedalWalker-v2').instantiate(seed=seed)
# Again, use a helper to create a model
# But because model is owned by the reinforcer, model should not be accessed using this variable
# but from reinforcer.model property
model_factory = DeterministicPolicyModelFactory(
policy_backbone=MLPFactory(input_length=24,
hidden_layers=[64, 64],
normalization='layer'),
value_backbone=MLPFactory(input_length=28,
hidden_layers=[64, 64],
normalization='layer'))
# Reinforcer - an object managing the learning process
reinforcer = BufferedSingleOffPolicyIterationReinforcer(
device=device,
settings=BufferedSingleOffPolicyIterationReinforcerSettings(
batch_rollout_rounds=4,
batch_training_rounds=1,
batch_size=32,
discount_factor=0.99),
environment=env,
algo=DeepDeterministicPolicyGradient(model_factory=model_factory,
tau=0.01,
max_grad_norm=0.5),
model=model_factory.instantiate(action_space=env.action_space),
env_roller=DequeReplayRollerOuNoise(environment=env,
device=device,
batch_size=32,
buffer_capacity=100,
buffer_initial_size=100,
noise_std_dev=0.2,
normalize_observations=True,
discount_factor=0.99))
# Model optimizer
optimizer = optim.Adam(reinforcer.model.parameters(), lr=2.5e-4, eps=1e-4)
# Overall information store for training information
training_info = TrainingInfo(
metrics=[
EpisodeRewardMetric(
'episode_rewards'), # Calculate average reward from episode
],
callbacks=[FrameTracker(100_000)
] # Print live metrics every epoch to standard output
)
# A bit of training initialization bookkeeping...
training_info.initialize()
reinforcer.initialize_training(training_info)
training_info.on_train_begin()
# Let's make 100 batches per epoch to average metrics nicely
num_epochs = 1
# Normal handrolled training loop
for i in range(1, num_epochs + 1):
epoch_info = EpochInfo(training_info=training_info,
global_epoch_idx=i,
batches_per_epoch=1,
optimizer=optimizer)
reinforcer.train_epoch(epoch_info, interactive=False)
training_info.on_train_end()
def test_prioritized_dqn_breakout():
"""
Simple 1 iteration of DQN prioritized replay breakout
"""
device = torch.device('cpu')
seed = 1001
# Set random seed in python std lib, numpy and pytorch
set_seed(seed)
# Only single environment for DQN
env = ClassicAtariEnv('BreakoutNoFrameskip-v4').instantiate(seed=seed)
# Again, use a helper to create a model
# But because model is owned by the reinforcer, model should not be accessed using this variable
# but from reinforcer.model property
model_factory = QModelFactory(backbone=NatureCnnFactory(
input_width=84, input_height=84, input_channels=4))
# Reinforcer - an object managing the learning process
reinforcer = BufferedSingleOffPolicyIterationReinforcer(
device=device,
settings=BufferedSingleOffPolicyIterationReinforcerSettings(
batch_rollout_rounds=4,
batch_training_rounds=1,
batch_size=32,
discount_factor=0.99),
environment=env,
algo=DeepQLearning(model_factory=model_factory,
double_dqn=False,
target_update_frequency=10_000,
max_grad_norm=0.5),
model=model_factory.instantiate(action_space=env.action_space),
env_roller=PrioritizedReplayRollerEpsGreedy(
environment=env,
device=device,
epsilon_schedule=LinearAndConstantSchedule(
initial_value=1.0, final_value=0.1, end_of_interpolation=0.1),
batch_size=8,
buffer_capacity=100,
priority_epsilon=1.0e-6,
buffer_initial_size=100,
frame_stack=4,
priority_exponent=0.6,
priority_weight=LinearSchedule(initial_value=0.4, final_value=1.0),
),
)
# Model optimizer
optimizer = optim.RMSprop(reinforcer.model.parameters(),
lr=2.5e-4,
alpha=0.95,
momentum=0.95,
eps=1e-3)
# Overall information store for training information
training_info = TrainingInfo(
metrics=[
EpisodeRewardMetric(
'episode_rewards'), # Calculate average reward from episode
],
callbacks=[FrameTracker(100_000)
] # Print live metrics every epoch to standard output
)
# A bit of training initialization bookkeeping...
training_info.initialize()
reinforcer.initialize_training(training_info)
training_info.on_train_begin()
# Let's make 100 batches per epoch to average metrics nicely
num_epochs = 1
# Normal handrolled training loop
for i in range(1, num_epochs + 1):
epoch_info = EpochInfo(training_info=training_info,
global_epoch_idx=i,
batches_per_epoch=1,
optimizer=optimizer)
reinforcer.train_epoch(epoch_info, interactive=False)
training_info.on_train_end()
def qbert_ppo():
device = torch.device('cuda:0')
seed = 1001
# Set random seed in python std lib, numpy and pytorch
set_seed(seed)
# Create 16 environments evaluated in parallel in sub processess with all usual DeepMind wrappers
# These are just helper functions for that
vec_env = SubprocVecEnvWrapper(ClassicAtariEnv('QbertNoFrameskip-v4'),
frame_history=4).instantiate(
parallel_envs=8, seed=seed)
# Again, use a helper to create a model
# But because model is owned by the reinforcer, model should not be accessed using this variable
# but from reinforcer.model property
model = StochasticPolicyModelFactory(
input_block=ImageToTensorFactory(),
backbone=NatureCnnFactory(
input_width=84, input_height=84,
input_channels=4)).instantiate(action_space=vec_env.action_space)
# Set schedule for gradient clipping.
cliprange = LinearSchedule(initial_value=0.1, final_value=0.0)
# Reinforcer - an object managing the learning process
reinforcer = OnPolicyIterationReinforcer(
device=device,
settings=OnPolicyIterationReinforcerSettings(batch_size=256,
experience_replay=4,
number_of_steps=128),
model=model,
algo=PpoPolicyGradient(entropy_coefficient=0.01,
value_coefficient=0.5,
max_grad_norm=0.5,
discount_factor=0.99,
gae_lambda=0.95,
cliprange=cliprange),
env_roller=StepEnvRoller(
environment=vec_env,
device=device,
))
# Model optimizer
optimizer = optim.Adam(reinforcer.model.parameters(),
lr=2.5e-4,
eps=1.0e-5)
# Overall information store for training information
training_info = TrainingInfo(
metrics=[
EpisodeRewardMetric(
'episode_rewards'), # Calculate average reward from episode
],
callbacks=[
StdoutStreaming(
), # Print live metrics every epoch to standard output
FrameTracker(
1.1e7
) # We need frame tracker to track the progress of learning
])
# A bit of training initialization bookkeeping...
training_info.initialize()
reinforcer.initialize_training(training_info)
training_info.on_train_begin()
# Let's make 10 batches per epoch to average metrics nicely
# Rollout size is 8 environments times 128 steps
num_epochs = int(1.1e7 / (128 * 8) / 10)
# Normal handrolled training loop
for i in range(1, num_epochs + 1):
epoch_info = EpochInfo(training_info=training_info,
global_epoch_idx=i,
batches_per_epoch=10,
optimizer=optimizer)
reinforcer.train_epoch(epoch_info)
training_info.on_train_end()