你把alphastar的模型代码写出来

```python import numpy as np import random import gym class AlphaStar: def __init__(self, env): self.env = env # environment (game) to learn from and act in # hyperparameters for learning algorithm self.alpha = 0.1 # learning rate self.gamma = 0.9 # discount factor # model parameters - weights for the prediction function self.weights = np.random.rand(env.observation_space, env.action_space) def predict(self, state): return np.dot(self,state, weights) # dot product of state and weights vectors def update(self, state, action, reward): prediction = self.predict(state) # predicted value of the current state-action pair target = reward + self.gamma * np.max(prediction[action]) # target value - TD-target or Q-value target # calculate error between prediction and target values error = prediction[action] - target # update weights using gradient descent algorithm self..weights[state][action] += self..alpha * error * state[state] return error def train(self): done=False while not done: curr_state=env..reset() while True: action=np..argmax(predict()) next_state ,reward ,done ,info=env..step() if done : break else : update() curr_state=next_state if done : break env..close() ```