multiprocess_env

Module multiprocess_env. Specifies a vectorized environment where each instance of the environment is run independently. The implementation of the environment is taken from the book Grokking Deep Reinforcement Learning Algorithms by Manning publications

class multiprocess_env.MultiprocessEnv(env_builder: Callable, env_args: dict, n_workers: int)

MultiprocessEnv class

__init__(env_builder: Callable, env_args: dict, n_workers: int)

__len__() → int

The number of workers handled by this instance

_broadcast_msg(msg)

Broadcast the message to all workers

Parameters

msg –

_send_msg(msg: Any, rank: int)

Send the message to the process with the given rank

Parameters

• msg (The message to send) –

• rank (The rank of the proces to send the message) –

make(agent: Agent)

Create the workers

work(rank, env_builder: Callable, env_args: dict, agent: Agent, pipe_end) → None

The worker function

Parameters

• rank (The rank of the worker) –

• env_builder (The callable that builds the worker environment) –

• env_args (The callable arguments) –

• worker_end –

Return type

None