return
self.body.memory.update(state, action, reward, next_state, done)
loss = self.algorithm.train()
if not np.isnan(loss): // set for log_summary()
self.body.loss = loss
explore_var = self.algorithm.update()
return loss, explore_var
@lab_api
def save(self, ckpt=None):
"""Save agent"""
if util.in_eval_lab_modes(): // eval does not save new models
return
self.algorithm.save(ckpt=ckpt)