Shortcuts

DdpgMlpActor

class torchrl.modules.DdpgMlpActor(action_dim: int, mlp_net_kwargs: Optional[dict] = None, device: Optional[Union[device, str, int]] = None)[source]

DDPG Actor class.

Presented in “CONTINUOUS CONTROL WITH DEEP REINFORCEMENT LEARNING”, https://arxiv.org/pdf/1509.02971.pdf

The DDPG Actor takes as input an observation vector and returns an action from it. It is trained to maximise the value returned by the DDPG Q Value network.

Parameters:
  • action_dim (int) – length of the action vector

  • mlp_net_kwargs (dict, optional) –

    kwargs for MLP. Defaults to

    >>> {
    ...     'in_features': None,
    ...     'out_features': action_dim,
    ...     'depth': 2,
    ...     'num_cells': [400, 300],
    ...     'activation_class': nn.ELU,
    ...     'bias_last_layer': True,
    ... }
    

  • device (torch.device, optional) – device to create the module on.

Examples

>>> import torch
>>> from torchrl.modules import DdpgMlpActor
>>> actor = DdpgMlpActor(action_dim=4)
>>> print(actor)
DdpgMlpActor(
  (mlp): MLP(
    (0): LazyLinear(in_features=0, out_features=400, bias=True)
    (1): ELU(alpha=1.0)
    (2): Linear(in_features=400, out_features=300, bias=True)
    (3): ELU(alpha=1.0)
    (4): Linear(in_features=300, out_features=4, bias=True)
  )
)
>>> obs = torch.zeros(10, 6)
>>> action = actor(obs)
>>> print(action.shape)
torch.Size([10, 4])
forward(observation: Tensor) Tensor[source]

Define the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

Docs

Access comprehensive developer documentation for PyTorch

View Docs

Tutorials

Get in-depth tutorials for beginners and advanced developers

View Tutorials

Resources

Find development resources and get your questions answered

View Resources