Features¶
Multi-input and non-sequential network architectures¶
Multi-input and other non-sequential networks are specified as list of layer lists, as opposed to
simply a list of layers for sequential networks. The following example illustrates how to specify
such a more complex network, by using the special layers Register
and Retrieve
to combine the
multiple sequential layer stacks.
Agent.create(
states=dict(
observation=dict(type='float', shape=(16, 16, 3)),
attributes=dict(type='int', shape=(4, 2), num_values=5)
),
...
policy=[
[
dict(type='retrieve', tensors='observation'),
dict(type='conv2d', size=16),
dict(type='flatten'),
dict(type='register', tensor='obs-embedding')
],
[
dict(type='retrieve', tensors='attributes'),
dict(type='embedding', size=16),
dict(type='flatten'),
dict(type='register', tensor='attr-embedding')
],
[
dict(
type='retrieve', tensors=['obs-embedding', 'attr-embedding'],
aggregation='concat'
),
dict(type='dense', size=32)
]
],
...
)
Action masking¶
agent = Agent.create(
states=dict(type='float', shape=(10,)),
actions=dict(type='int', shape=(), num_actions=3), ...
)
...
states = dict(
state=np.random.random_sample(size=(10,)), # regular state
action_mask=[True, False, True] # mask as '[ACTION-NAME]_mask'
)
action = agent.act(states=states)
assert action != 1
Parallel environment execution¶
Execute multiple environments running locally in one call / batched:
Runner(
agent='benchmarks/configs/ppo1.json', environment='CartPole-v1',
num_parallel=5
)
runner.run(num_episodes=100, batch_agent_calls=True)
Execute environments running in different processes whenever ready / unbatched:
Runner(
agent='benchmarks/configs/ppo1.json', environment='CartPole-v1',
num_parallel=5, remote='multiprocessing'
)
runner.run(num_episodes=100)
Execute environments running on different machines, here using run.py
instead
of Runner
:
# Environment machine 1
python run.py --environment gym --level CartPole-v1 --remote socket-server \
--port 65432
# Environment machine 2
python run.py --environment gym --level CartPole-v1 --remote socket-server \
--port 65433
# Agent machine
python run.py --agent benchmarks/configs/ppo1.json --episodes 100 \
--num-parallel 2 --remote socket-client --host 127.0.0.1,127.0.0.1 \
--port 65432,65433 --batch-agent-calls
Record & pretrain¶
agent = Agent.create(...
recorder=dict(
directory='data/traces',
frequency=100 # record a traces file every 100 episodes
), ...
)
...
agent.close()
# Pretrain agent on recorded traces
agent = Agent.create(...)
agent.pretrain(
directory='data/traces',
num_iterations=100 # perform 100 update iterations on traces (more configurations possible)
)
Save & restore¶
TensorFlow saver (full model)¶
agent = Agent.create(...
saver=dict(
directory='data/checkpoints',
frequency=600 # save checkpoint every 600 seconds (10 minutes)
), ...
)
...
agent.close()
# Restore latest agent checkpoint
agent = Agent.load(directory='data/checkpoints')
NumPy / HDF5 (only weights)¶
agent = Agent.create(...
saver=dict(
directory='data/checkpoints',
frequency=600 # save checkpoint every 600 seconds (10 minutes)
), ...
)
...
agent.save(directory='data/checkpoints', format='numpy', append='episodes')
# Restore latest agent checkpoint
agent = Agent.load(directory='data/checkpoints', format='numpy')
TensorBoard¶
Agent.create(...
summarizer=dict(
directory='data/summaries',
# list of labels, or 'all'
labels=['graph', 'entropy', 'kl-divergence', 'losses', 'rewards'],
frequency=100, # store values every 100 timesteps
# (infrequent update summaries every update; other configurations possible)
custom=dict( # custom summaries which need to be recorded explicitly
custom_summary1=dict(type='image', max_outputs=10), ...
)
), ...
)
...
# custom summary recording
agent.summarize(summary='custom_summary1', value=image)