agent.reset() not called between subsequent episodes in sim2real challenge

[karkuspeter]

Hi,

One would expect agent.reset() to be called in this loop:

iGibson/gibson2/envs/challenge.py

Lines 22 to 32 in 3453531

	for i in range(num_eval_episodes):
	print('Episode: {}/{}'.format(i + 1, num_eval_episodes))
	state = self.nav_env.reset()
	while True:
	action = agent.act(state)
	state, reward, done, info = self.nav_env.step(action)
	total_reward += reward
	if done:
	break
	total_success += info['success']
	total_spl += info['spl']

Is it missing by mistake?

[ChengshuLi]

Thank you for pointing this out. We have added agent.reset() in this comment c96cef9.

This will be especially useful for recurrent policies because you would need to reset the hidden state in agent.reset().

Feel free to re-open this issue if you have any other questions.