salina.agents.gyma

salina.agents.gyma.GymAgent (TAgent)

Create an Agent from a gyn environment

__init__(self, make_env_fn=None, make_env_args={}, n_envs=None, input='action', output='env/', use_seed=True) special

Create an agent from a Gym environment

Parameters:

Name	Type	Description	Default
make_env_fn	[function that returns a gym.Env]	The function to create a single gym environments	None
make_env_args	dict	The arguments of the function that creates a gym.Env	{}
n_envs	[int]	The number of environments to create.	None
input	str	[the name of the action variable in the workspace]. Defaults to "action".	'action'
output	str	[the output prefix of the environment]. Defaults to "env/".	'env/'
use_seed	bool	[If True, then the seed is chained to the environments, and each environment will have its own seed]. Defaults to True.	True

Source code in salina/agents/gyma.py

def __init__(
    self,
    make_env_fn=None,
    make_env_args={},
    n_envs=None,
    input="action",
    output="env/",
    use_seed=True
):
    """ Create an agent from a Gym environment

    Args:
        make_env_fn ([function that returns a gym.Env]): The function to create a single gym environments
        make_env_args (dict): The arguments of the function that creates a gym.Env
        n_envs ([int]): The number of environments to create.
        input (str, optional): [the name of the action variable in the workspace]. Defaults to "action".
        output (str, optional): [the output prefix of the environment]. Defaults to "env/".
        use_seed (bool, optional): [If True, then the seed is chained to the environments, and each environment will have its own seed]. Defaults to True.
    """
    super().__init__()
    self.use_seed=use_seed
    assert n_envs > 0
    self.envs = None
    self.env_args = make_env_args
    self._seed = 0
    self.n_envs = n_envs
    self.output = output
    self.input = input
    self.make_env_fn = make_env_fn
    self.ghost_params = torch.nn.Parameter(torch.randn(()))

salina.agents.gyma.GymAgent (TAgent)

Create an Agent from a gyn environment

__init__(self, make_env_fn=None, make_env_args={}, n_envs=None, input='action', output='env/', use_seed=True) special

Create an agent from a Gym environment

Parameters:

Name	Type	Description	Default
make_env_fn	[function that returns a gym.Env]	The function to create a single gym environments	None
make_env_args	dict	The arguments of the function that creates a gym.Env	{}
n_envs	[int]	The number of environments to create.	None
input	str	[the name of the action variable in the workspace]. Defaults to "action".	'action'
output	str	[the output prefix of the environment]. Defaults to "env/".	'env/'
use_seed	bool	[If True, then the seed is chained to the environments, and each environment will have its own seed]. Defaults to True.	True

Source code in salina/agents/gyma.py

def __init__(
    self,
    make_env_fn=None,
    make_env_args={},
    n_envs=None,
    input="action",
    output="env/",
    use_seed=True
):
    """ Create an agent from a Gym environment

    Args:
        make_env_fn ([function that returns a gym.Env]): The function to create a single gym environments
        make_env_args (dict): The arguments of the function that creates a gym.Env
        n_envs ([int]): The number of environments to create.
        input (str, optional): [the name of the action variable in the workspace]. Defaults to "action".
        output (str, optional): [the output prefix of the environment]. Defaults to "env/".
        use_seed (bool, optional): [If True, then the seed is chained to the environments, and each environment will have its own seed]. Defaults to True.
    """
    super().__init__()
    self.use_seed=use_seed
    assert n_envs > 0
    self.envs = None
    self.env_args = make_env_args
    self._seed = 0
    self.n_envs = n_envs
    self.output = output
    self.input = input
    self.make_env_fn = make_env_fn
    self.ghost_params = torch.nn.Parameter(torch.randn(()))