training issue

[ShelyH]

Hi,sorry to bother you.
I don't want to imitate learning,so I set 'imitation_learning=False' in train.py,but there are some error:
AttributeError: 'JointState' object has no attribute 'unsqueeze' File "train.py", line 188, in
main()
File "train.py", line 140, in main
explorer.run_k_episodes(il_episodes, 'train', update_memory=True, imitation_learning=False)
File "../crowd_nav/utils/explorer.py", line 72, in run_k_episodes
self.update_memory(states, actions, rewards, imitation_learning)
File "../crowd_nav/utils/explorer.py", line 116, in update_memory
value = reward + gamma_bar * self.target_model(next_state.unsqueeze(0)).data.item()
AttributeError: 'JointState' object has no attribute 'unsqueeze'

Can you tell me how to solve it?Thanks very much!

[ChanganVR]

@2529342549 if you don't want to run imitation learning, you should comment out these lines altogether instead of just setting that variable to False:

CrowdNav/crowd_nav/train.py

Lines 116 to 133 in 20d6780

	il_episodes = train_config.getint('imitation_learning', 'il_episodes')
	il_policy = train_config.get('imitation_learning', 'il_policy')
	il_epochs = train_config.getint('imitation_learning', 'il_epochs')
	il_learning_rate = train_config.getfloat('imitation_learning', 'il_learning_rate')
	trainer.set_learning_rate(il_learning_rate)
	if robot.visible:
	safety_space = 0
	else:
	safety_space = train_config.getfloat('imitation_learning', 'safety_space')
	il_policy = policy_factory[il_policy]()
	il_policy.multiagent_training = policy.multiagent_training
	il_policy.safety_space = safety_space
	robot.set_policy(il_policy)
	explorer.run_k_episodes(il_episodes, 'train', update_memory=True, imitation_learning=True)
	trainer.optimize_epoch(il_epochs)
	torch.save(model.state_dict(), il_weight_file)
	logging.info('Finish imitation learning. Weights saved.')
	logging.info('Experience set size: %d/%d', len(memory), memory.capacity)

Let me know if this fix works for you.

[ShelyH]

Hi, thanks for your last answer，I'm here to disturb you again。
During training, the model is evaluated every 1000 intervals，the purpose of this is to show how good or bad the current model is or if there are other intentions？
# evaluate the model if episode % evaluation_interval == 0: explorer.run_k_episodes(env.case_size['val'], 'val', episode=episode)