Carla simulator with RL 该分支主要是用于跑baseline，和它做一个比较，参数如下所示： "args":[ "--root_dir", "logs/main/", "--experiment_name", "latent_sac", "--gin_file", "params.gin", "--gin_param", "load_carla_env.port=2000" ], "env": { "CUDA_VISIBLE_DEVICES":"0", "PYTHONPATH":"${workspaceRoot}" }
If out of system memory, change the parameter replay_buffer_capacity and initial_collect_steps the function tran_eval smaller.
If out of CUDA memory, set parameter model_batch_size or sequence_length of the function tran_eval smaller.