Source code for alf.networks.action_encoder

# Copyright (c) 2019 Horizon Robotics. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#      http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""A simple parameterless action encoder."""

import numpy as np
import torch
import torch.nn.functional as F

import alf

from .network import Network


class SimpleActionEncoder(Network):
    """A simple encoder for action.

    It encodes discrete action to one hot representation and use the original
    continous actions. The output is the concat of all of them after flattening.
    """

    def __init__(self, action_spec):
        """

        Args:
            action_spec (nested BoundedTensorSpec): spec for actions
        """

        def check_supported_spec(spec):
            if spec.is_discrete:
                assert np.min(spec.minimum) == np.max(spec.minimum) == 0
                assert np.min(spec.maximum) == np.max(spec.maximum)

        alf.nest.map_structure(check_supported_spec, action_spec)
        self._action_spec = action_spec
        super().__init__(input_tensor_spec=action_spec, name="ActionEncoder")

    def forward(self, inputs, state=()):
        """Generate encoded actions.

        Args:
            inputs (nested Tensor): action tensors.
        Returns:
            nested Tensor with the same structure as inputs.

        """
        alf.nest.assert_same_structure(inputs, self._action_spec)
        actions = inputs
        outer_rank = alf.nest.utils.get_outer_rank(inputs, self._action_spec)

        def _encode_one_action(action, spec):
            if spec.is_discrete:
                num_actions = spec.maximum - spec.minimum + 1
                if num_actions.ndim == 0:
                    num_actions = int(num_actions)
                else:
                    num_actions = int(num_actins[0])
                a = F.one_hot(action, num_actions).to(torch.float32)
            else:
                a = action
            if outer_rank > 0:
                return a.reshape(*a.shape[:outer_rank], -1)
            else:
                return a.reshape(-1)

        actions = alf.nest.map_structure(_encode_one_action, actions,
                                         self._action_spec)

        return torch.cat(alf.nest.flatten(actions), dim=-1), ()