some confusion about Evaluator #462
Description
May I ask why the author is in the class Evaluator:
def init(self, cwd: str, env, args: Config, if_tensorboard: bool = False):
self.cwd = cwd # current working directory to save model
self.env = env # the env for Evaluator, eval_env = env in default
self.agent_id = args.gpu_id
self.total_step = 0 # the total training step
self.start_time = time.time() # used_time = time.time() - self.start_time
self.eval_times = args.eval_times # number of times that get episodic cumulative return
self.eval_per_step = args.eval_per_step # evaluate the agent per training steps
self.eval_step_counter = -self.eval_per_step # self.total_step > self.eval_step_counter + self.eval_per_step
Here self.eval_step_counter= -self.eval_per_step is selected and self.eval_step_counter=0 in hellow dqn