Description
What happened + What you expected to happen
When i tried to use SAC to train my model, the error happened.
Versions / Dependencies
Ray:Version: 2.46.0
python:Python 3.10.17
Reproduction script
agent, checkpoint_path = load_policy(algo, env_name, load_policy_path, env_config=env_config, seed=seed)
def load_policy(algo, env_name, policy_path=None, seed=0, env_config={}, eval=False):
if algo == 'ppo':
agent = ppo.PPOTrainer(setup_config(algo, seed, env_config, eval=eval), env_name,
logger_creator=custom_log_creator("", env_name)
)
elif algo == 'sac':
config = setup_config(algo, seed, env_config, eval=eval)
agent = SAC(config=config,
logger_creator=custom_log_creator("", env_name)
)
When i tried to use agent = SAC(config=config, logger_creator=custom_log_creator("**", env_name), the error happened. The EnvRunnerGroup's setup method self._local_config = local_config_raw.framework(tf_session_args=config.copy(copy_frozen=False)). The return error TypeError: 'str' object is not callable. However my config type is <class 'ray.rllib.algorithms.sac.sac.SACConfig'>
Issue Severity
High: It blocks me from completing my task.