solve_cart_pole.py 文件源码

def train_on_memory(model, memory_container, episode):
    # container contains [(observation0, rewards_vec0), (observation1, rewards_vec1),...]
    # x contains model fits X which is the states, y contains model fits y which is the rewards_vec in every state
    x = np.array(memory_container['observations'])
    y = np.array(memory_container['rewards'])

    early_stopping = EarlyStopping(monitor='val_loss', min_delta=0, verbose=1, patience=2)
    model.fit(x, y, epochs=3000, validation_split=0.3, callbacks=[early_stopping])
    return model