Deepair implementations of reinforcement learning algorithms. It focus on DRL algorithms and implementing the latest advancements in DRL. Highly customizable support for training processes. Suitable for the research and application of the latest technologies in reinforcement learning.
Documentation is available: https://deepair.readthedocs.io/
pip install deepair
or
pip install git+https://github.com/sonnhfit/deepair.git
import gym
from deepair.dqn import Rainbow
env = gym.make('LunarLander-v2')
rain = Rainbow(env=env, memory_size=10000, batch_size=32, target_update=256)
rain.train(timesteps=200000)
# test
state = env.reset()
done = False
score = 0
while not done:
action = rain.select_action(state, deterministic=True)
next_state, reward, done, info = env.step(action)
state = next_state
score += reward
print("score: ", score)
- save model
- load model