I want to understand data stream in Tianshou. How the data is pre-processed or we must do it ourselves?

I have read the example here. Tic tac too PettingZoo enviroment have Dict space observation. Observation contains key ‘observation’ and key ‘action_mask’:

Dict('action_mask': Box(0, 1, (9,), int8), 'observation': Box(0, 1, (3, 3, 2), int8))

But policy have only model for ‘observation’ (input dim = 18 = 3x3x2):

DQNPolicy(
  (model): Net(
    (model): MLP(
      (model): Sequential(
        (0): Linear(in_features=18, out_features=128, bias=True)
        (1): ReLU()
        (2): Linear(in_features=128, out_features=128, bias=True)
        (3): ReLU()
        (4): Linear(in_features=128, out_features=128, bias=True)
        (5): ReLU()
        (6): Linear(in_features=128, out_features=128, bias=True)
        (7): ReLU()
        (8): Linear(in_features=128, out_features=9, bias=True)
      )
    )
  )
  (model_old): Net(
    (model): MLP(
      (model): Sequential(
        (0): Linear(in_features=18, out_features=128, bias=True)
        (1): ReLU()
        (2): Linear(in_features=128, out_features=128, bias=True)
        (3): ReLU()
        (4): Linear(in_features=128, out_features=128, bias=True)
        (5): ReLU()
        (6): Linear(in_features=128, out_features=128, bias=True)
        (7): ReLU()
        (8): Linear(in_features=128, out_features=9, bias=True)
      )
    )
  )
)

I tried to understand it by looking at the source code but I couldn’t find it. Do anyone can explain me? Thanks.