pyhanabi/belief_model.py [89:113]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        self.net = nn.Sequential(
            nn.Linear(self.in_dim, self.hid_dim),
            nn.ReLU(),
            nn.Linear(self.hid_dim, self.hid_dim),
            nn.ReLU(),
        )
        self.lstm = nn.LSTM(
            self.hid_dim,
            self.hid_dim,
            num_layers=self.num_lstm_layer,
        ).to(device)
        self.lstm.flatten_parameters()

        self.emb = nn.Linear(25, self.hid_dim // 8, bias=False)
        self.auto_regress = nn.LSTM(
            self.hid_dim + self.hid_dim // 8,
            self.hid_dim,
            num_layers=1,
            batch_first=True,
        ).to(device)
        self.auto_regress.flatten_parameters()

        self.fc = nn.Linear(self.hid_dim, self.out_dim)

    @torch.jit.script_method
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



pyhanabi/tools/convert_model.py [108:132]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        self.net = nn.Sequential(
            nn.Linear(self.in_dim, self.hid_dim),
            nn.ReLU(),
            nn.Linear(self.hid_dim, self.hid_dim),
            nn.ReLU(),
        )
        self.lstm = nn.LSTM(
            self.hid_dim,
            self.hid_dim,
            num_layers=self.num_lstm_layer,
        ).to(device)
        self.lstm.flatten_parameters()

        self.emb = nn.Linear(25, self.hid_dim // 8, bias=False)
        self.auto_regress = nn.LSTM(
            self.hid_dim + self.hid_dim // 8,
            self.hid_dim,
            num_layers=1,
            batch_first=True,
        ).to(device)
        self.auto_regress.flatten_parameters()

        self.fc = nn.Linear(self.hid_dim, self.out_dim)

    @torch.jit.script_method
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -