How should I load the model which is saved while training of cartpole game. so that i can just play the game (testing) without further training
I’ve tried this code:
lodded_model = Agent(4,2)
lodded_model.load(“cartpole_model/weights_350.hdf5”)
for e in range(30):
observation = env.reset()
observation = observation.reshape((1,state_size))
for t in range(500):
env.render()
action = lodded_model.act(observation)
new_observation, reward, done, info = env.step(action)
new_observation = new_observation.reshape((1,state_size))
observation = new_observation
if done:
print("Episode {}/{} with score {}".format(e,30,t))
break
env.close()
print(“All episodes done”)`
But this gives, random behaviour and low score, whereas if i run this action = agent.act(observation) it performs well. why new model is not able to adapt all those things as that of agent model