latent_sac

[gacdan]

Hello, I am very interested in your research. During the training, I have some questions as follows :
1、 when agent_name=='latent_sac',I could not find the 'tf_agent.train_model()' function in line 740 in the train_eval.py, should it be revised to 'tf agent.train ()'?
2、I'm not sure which output interface is used to interact with the carla environment and control the vehicle during the training process. I understand you mean the function 'collect_driver=dynamic_step_driver.DynamicStepDriver()' is the output interface in line 681 in your train_eval.py, but when agent_name=='latent_sac' in the later training in line 752 in your train_eval.py , I could not see the 'collect_driver.run()' function, Does it means that the 'step(action)' operation in carla environment is not performed in the training process, and could not control the vehicle?
We look forward to your prompt response, Thank you !