adik993 / ppo-pytorch Goto Github PK

I really love this implementation, and I see that LSTM is still in the TODO. Have you made any progress on this in the last two months or should I just do it myself?

I tried your script in mountaincar env and It seems that the game ends when the step length reaches 200 per episode, but in your tensorboard plots, an episode didn't stop until it reached the final state(the top of mountain). I wonder if it's because there is any early ending mechanisms in your code but unfortunately I didn't find it. Could you give me some advise to get your tensorboard result in your publishes?

I want to train the agent using the project file after customizing the environment based on the gym's continuous space, the state and actions of the environment are defined as follows：

    self.min_action = np.array([[-3, -3, -3, -3, -3]]).reshape(1,5)
    self.max_action = np.array([[3, 3, 3, 3, 3]]).reshape(1,5)

    self.low_state = np.array(
        [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], dtype=np.float32
    ).reshape(1,10)
    self.high_state = np.array(
        [[50, 50, 50, 50, 50, 300, 300, 300, 300, 300]], dtype=np.float32
    ).reshape(1,10)

    self.action_space = spaces.Box(
        low=self.min_action, high=self.max_action, shape=(1, 5), dtype=np.float32
    )

    self.observation_space = spaces.Box(
        low=self.low_state, high=self.high_state, shape=(1, 10), dtype=np.float32
    )

Is it possible to implement this idea based on PPO ICM? Thanks!

I tried your run_mountain_car.py, but the accumulated rewards do not change at all.
Are there any hyper-parameters that I need to change? And how many episodes are needed in general?