def forward()

in src_code/modules/rnn_interactive_agent.py [0:0]

• 23 lines of code
• 2 McCabe index (conditional complexity)

    def forward(self, inputs, inputs_alone, hidden_state, hidden_state_alone, hidden_state_):
        _inputs = inputs.clone()
        if self.args.obs_agent_id:
            bs = _inputs.shape[0]
            # World features
            world_feats = _inputs[:, :-self.individual_feats_size-self.n_agents-self.args.n_actions-self.all_feats_size*(self.n_agents-1)]
            action_id_feats = _inputs[:, -self.n_agents-self.args.n_actions:]
            self_feats = _inputs[:, -self.n_agents-self.args.n_actions-self.individual_feats_size:-self.n_agents-self.args.n_actions]
            ally_feats = _inputs[:, -self.individual_feats_size-self.n_agents-self.args.n_actions-self.all_feats_size*(self.n_agents-1):-self.n_agents-self.args.n_actions-self.individual_feats_size].reshape(bs, self.n_agents-1, -1)
            _inputs = th.cat((world_feats, th.zeros(ally_feats.reshape(bs, -1).shape, device=self.args.device), self_feats, action_id_feats), dim=-1)
        else:
            bs = _inputs.shape[0]
            # World features
            world_feats = _inputs[:, :-self.individual_feats_size-self.args.n_actions-self.all_feats_size*(self.n_agents-1)]
            action_id_feats = _inputs[:, -self.args.n_actions:]
            self_feats = _inputs[:, -self.args.n_actions-self.individual_feats_size:-self.args.n_actions]
            ally_feats = _inputs[:, -self.individual_feats_size-self.args.n_actions-self.all_feats_size*(self.n_agents-1):-self.args.n_actions-self.individual_feats_size].reshape(bs, self.n_agents-1, -1)
            _inputs = th.cat((world_feats, th.zeros(ally_feats.reshape(bs, -1).shape, device=self.args.device), self_feats, action_id_feats), dim=-1)

        inputs = self.self_attn_i(inputs)
        _inputs = self.self_attn_i(_inputs)
        h_alone, q_alone = self.agent_alone(inputs_alone, hidden_state_alone)
        h_interactive_, q_interactive_ = self.agent_interactive(_inputs, hidden_state_)
        h_interactive, q_interactive = self.agent_interactive(inputs, hidden_state)

        q = q_alone + q_interactive
        return q, h_interactive, h_alone, h_interactive_, q_interactive_