File size: 12,786 Bytes
079c32c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
import copy
from typing import TYPE_CHECKING, List, Any, Union

import numpy as np
import torch
from easydict import EasyDict

from lzero.policy import InverseScalarTransform
from lzero.mcts.ctree.ctree_stochastic_muzero import stochastic_mz_tree


# ==============================================================
# Stochastic MuZero
# ==============================================================


class StochasticMuZeroMCTSCtree(object):
    """
    Overview:
        MCTSCtree for Stochastic MuZero. The core ``batch_traverse`` and ``batch_backpropagate`` function is implemented in C++.

    Interfaces:
        __init__, roots, search
    """

    config = dict(
        # (float) The alpha value used in the Dirichlet distribution for exploration at the root node of the search tree.
        root_dirichlet_alpha=0.3,
        # (float) The noise weight at the root node of the search tree.
        root_noise_weight=0.25,
        # (int) The base constant used in the PUCT formula for balancing exploration and exploitation during tree search.
        pb_c_base=19652,
        # (float) The initialization constant used in the PUCT formula for balancing exploration and exploitation during tree search.
        pb_c_init=1.25,
        # (float) The maximum change in value allowed during the backup step of the search tree update.
        value_delta_max=0.01,
    )

    @classmethod
    def default_config(cls: type) -> EasyDict:
        cfg = EasyDict(copy.deepcopy(cls.config))
        cfg.cfg_type = cls.__name__ + 'Dict'
        return cfg

    def __init__(self, cfg: EasyDict = None) -> None:
        """
        Overview:
            Use the default configuration mechanism. If a user passes in a cfg with a key that matches an existing key
            in the default configuration, the user-provided value will override the default configuration. Otherwise,
            the default configuration will be used.
        """
        default_config = self.default_config()
        default_config.update(cfg)
        self._cfg = default_config
        self.inverse_scalar_transform_handle = InverseScalarTransform(
            self._cfg.model.support_scale, self._cfg.device, self._cfg.model.categorical_distribution
        )

    @classmethod
    def roots(cls: int, active_collect_env_num: int, legal_actions: List[Any],
              chance_space_size: int = 2) -> "stochastic_mz_tree.Roots":
        """
        Overview:
            The initialization of CRoots with root num and legal action lists.
        Arguments:
            - root_num (:obj:`int`): the number of the current root.
            - legal_action_list (:obj:`list`): the vector of the legal action of this root.
        """
        from lzero.mcts.ctree.ctree_stochastic_muzero import stochastic_mz_tree as ctree
        return ctree.Roots(active_collect_env_num, legal_actions, chance_space_size)

    def search(
            self, roots: Any, model: torch.nn.Module, latent_state_roots: List[Any], to_play_batch: Union[int,
            List[Any]]
    ) -> None:
        """
        Overview:
            Do MCTS for the roots (a batch of root nodes in parallel). Parallel in model inference.
             Use the cpp ctree.
        Arguments:
            - roots (:obj:`Any`): a batch of expanded root nodes
            - latent_state_roots (:obj:`list`): the hidden states of the roots
            - to_play_batch (:obj:`list`): the to_play_batch list used in in self-play-mode board games
        """
        with torch.no_grad():
            model.eval()

            # preparation some constant
            batch_size = roots.num
            pb_c_base, pb_c_init, discount_factor = self._cfg.pb_c_base, self._cfg.pb_c_init, self._cfg.discount_factor
            # the data storage of latent states: storing the latent state of all the nodes in the search.
            latent_state_batch_in_search_path = [latent_state_roots]

            # minimax value storage
            min_max_stats_lst = stochastic_mz_tree.MinMaxStatsList(batch_size)
            min_max_stats_lst.set_delta(self._cfg.value_delta_max)

            for simulation_index in range(self._cfg.num_simulations):
                # In each simulation, we expanded a new node, so in one search, we have ``num_simulations`` num of nodes at most.

                latent_states = []

                # prepare a result wrapper to transport results between python and c++ parts
                results = stochastic_mz_tree.ResultsWrapper(num=batch_size)

                # latent_state_index_in_search_path: the first index of leaf node states in latent_state_batch_in_search_path, i.e. is current_latent_state_index in one the search.
                # latent_state_index_in_batch: the second index of leaf node states in latent_state_batch_in_search_path, i.e. the index in the batch, whose maximum is ``batch_size``.
                # e.g. the latent state of the leaf node in (x, y) is latent_state_batch_in_search_path[x, y], where x is current_latent_state_index, y is batch_index.
                # The index of value prefix hidden state of the leaf node are in the same manner.
                """
                MCTS stage 1: Selection
                    Each simulation starts from the internal root state s0, and finishes when the simulation reaches a leaf node s_l.
                """
                leaf_node_is_chance, latent_state_index_in_search_path, latent_state_index_in_batch, last_actions, virtual_to_play_batch = stochastic_mz_tree.batch_traverse(
                    roots, pb_c_base, pb_c_init, discount_factor, min_max_stats_lst, results,
                    copy.deepcopy(to_play_batch)
                )

                # obtain the latent state for leaf node
                for ix, iy in zip(latent_state_index_in_search_path, latent_state_index_in_batch):
                    latent_states.append(latent_state_batch_in_search_path[ix][iy])

                latent_states = torch.from_numpy(np.asarray(latent_states)).to(self._cfg.device).float()
                # .long() is only for discrete action
                last_actions = torch.from_numpy(np.asarray(last_actions)).to(self._cfg.device).long()
                """
                MCTS stage 2: Expansion
                    At the final time-step l of the simulation, the next_latent_state and reward/value_prefix are computed by the dynamics function.
                    Then we calculate the policy_logits and value for the leaf node (next_latent_state) by the prediction function. (aka. evaluation)
                MCTS stage 3: Backup
                    At the end of the simulation, the statistics along the trajectory are updated.
                """
                # network_output = model.recurrent_inference(latent_states, last_actions)

                num = len(leaf_node_is_chance)
                leaf_idx_list = list(range(num))
                latent_state_batch = [None] * num
                value_batch = [None] * num
                reward_batch = [None] * num
                policy_logits_batch = [None] * num
                child_is_chance_batch = [None] * num
                chance_nodes_index = []
                decision_nodes_index = []

                for i, leaf_node_is_chance_ in enumerate(leaf_node_is_chance):
                    if leaf_node_is_chance_:
                        chance_nodes_index.append(i)
                    else:
                        decision_nodes_index.append(i)

                def process_nodes(nodes_index, is_chance):
                    # Return early if nodes_index is empty
                    if not nodes_index:
                        return

                    # Slice and stack latent_states and last_actions based on nodes_index
                    latent_states_stack = torch.stack([latent_states[i] for i in nodes_index], dim=0)
                    last_actions_stack = torch.stack([last_actions[i] for i in nodes_index], dim=0)

                    # Pass the stacked batch through the recurrent_inference function
                    network_output_batch = model.recurrent_inference(latent_states_stack,
                                                                     last_actions_stack,
                                                                     afterstate=not is_chance)

                    # Split the batch output into separate nodes
                    latent_state_splits = torch.split(network_output_batch.latent_state, 1, dim=0)
                    value_splits = torch.split(network_output_batch.value, 1, dim=0)
                    reward_splits = torch.split(network_output_batch.reward, 1, dim=0)
                    policy_logits_splits = torch.split(network_output_batch.policy_logits, 1, dim=0)

                    for i, (latent_state, value, reward, policy_logits) in zip(nodes_index,
                                                                               zip(latent_state_splits, value_splits,
                                                                                   reward_splits,
                                                                                   policy_logits_splits)):
                        if not model.training:
                            value = self.inverse_scalar_transform_handle(value).detach().cpu().numpy()
                            reward = self.inverse_scalar_transform_handle(reward).detach().cpu().numpy()
                            latent_state = latent_state.detach().cpu().numpy()
                            policy_logits = policy_logits.detach().cpu().numpy()

                        latent_state_batch[i] = latent_state
                        value_batch[i] = value.reshape(-1).tolist()
                        reward_batch[i] = reward.reshape(-1).tolist()
                        policy_logits_batch[i] = policy_logits.tolist()
                        child_is_chance_batch[i] = is_chance

                process_nodes(chance_nodes_index, True)
                process_nodes(decision_nodes_index, False)
                chance_nodes = chance_nodes_index
                decision_nodes = decision_nodes_index

                value_batch_chance = [value_batch[leaf_idx] for leaf_idx in chance_nodes]
                value_batch_decision = [value_batch[leaf_idx] for leaf_idx in decision_nodes]
                reward_batch_chance = [reward_batch[leaf_idx] for leaf_idx in chance_nodes]
                reward_batch_decision = [reward_batch[leaf_idx] for leaf_idx in decision_nodes]
                policy_logits_batch_chance = [policy_logits_batch[leaf_idx] for leaf_idx in chance_nodes]
                policy_logits_batch_decision = [policy_logits_batch[leaf_idx] for leaf_idx in decision_nodes]

                latent_state_batch = np.concatenate(latent_state_batch, axis=0)
                latent_state_batch_in_search_path.append(latent_state_batch)

                # In ``batch_backpropagate()``, we first expand the leaf node using ``the policy_logits`` and
                # ``reward`` predicted by the model, then perform backpropagation along the search path to update the
                # statistics.

                # NOTE: simulation_index + 1 is very important, which is the depth of the current leaf node.
                current_latent_state_index = simulation_index + 1

                if (len(chance_nodes) > 0):
                    value_batch_chance = np.concatenate(value_batch_chance, axis=0).reshape(-1).tolist()
                    reward_batch_chance = np.concatenate(reward_batch_chance, axis=0).reshape(-1).tolist()
                    policy_logits_batch_chance = np.concatenate(policy_logits_batch_chance, axis=0).tolist()
                    stochastic_mz_tree.batch_backpropagate(
                        current_latent_state_index, discount_factor, reward_batch_chance, value_batch_chance,
                        policy_logits_batch_chance,
                        min_max_stats_lst, results, virtual_to_play_batch, child_is_chance_batch, chance_nodes
                    )
                if (len(decision_nodes) > 0):
                    value_batch_decision = np.concatenate(value_batch_decision, axis=0).reshape(-1).tolist()
                    reward_batch_decision = np.concatenate(reward_batch_decision, axis=0).reshape(-1).tolist()
                    policy_logits_batch_decision = np.concatenate(policy_logits_batch_decision, axis=0).tolist()
                    stochastic_mz_tree.batch_backpropagate(
                        current_latent_state_index, discount_factor, reward_batch_decision, value_batch_decision,
                        policy_logits_batch_decision,
                        min_max_stats_lst, results, virtual_to_play_batch, child_is_chance_batch, decision_nodes
                    )