[Bug Report] Bug title

[ChengYoghurt]

Describe the bug
I used the "CarRacing-v3" environment and set the "continuous=False", and created a stable baseline3 DQR agent to train on it, I got error message like this:

According to the error message, I checked the source code of "class CarRacing", and noticed that:

action = action.astype(np.float64)
will be executed even when action is sampled from Discrete Action Space, i.e. when action is an Integer.

Code example
`import gymnasium as gym
from gymnasium.spaces import Tuple, Discrete, Box
from stable_baselines3 import PPO, DQN
from stable_baselines3.common.monitor import Monitor
from stable_baselines3.common.vec_env import DummyVecEnv, VecVideoRecorder
import numpy as np

config = {
"policy_type": "MlpPolicy",
"total_timesteps": 25000,
"learning_rate": 0.01,
"gamma": 0.95,
"env_name": "CarRacing-v3",
}

env = gym.make(config["env_name"], render_mode="rgb_array", continuous=False,)
env = Monitor(env) # record stats such as returns
env = DummyVecEnv([make_env])

Initialize PPO model

model = DQN(
"MlpPolicy", # Use a multi-layer perceptron policy
env,
verbose=1,
learning_rate=config["learning_rate"],
# n_steps=512, # Number of steps to run for each environment per update
gamma=config["gamma"], # Discount factor
)

Train DQN

model.learn(total_timesteps=config["total_timesteps"], ),

`

System Info
Describe the characteristic of your environment:

• my Gym was installed via pip
• I ran into this problem on a linux machine
• Python version=3.12

Checklist

• [yes] I have checked that there is no similar issue in the repo (required)

[pseudo-rnd-thoughts]

Thank for raising this, it was fixed in Farama-Foundation/Gymnasium#1253 and will be part of Gymnasium v1.1

[ChengYoghurt]

I see. Thanks for your prompt help!