I create a DQNTrainer and load a checkpoint from a DQN agent that I trained. config = # loaded params.pkl config['evaluation_num_workers'] = 0 # I also tried with None, same result trainer = DQNTrainer(config=config) trainer.restore(checkpoint) trainer.config['evaluation_num_episodes'] = 1 metric…

It seems that this question was solved in this thread: [image] Cannot create R2D2 trainer with evaluation worker RLlib @fedetask , I would check what config["evaluation_config"] looks like as during evaluation this config is used. Maybe therein you get a default of 1. …

DQNTrainer evaluate() doesn't perform any episode

RLlib

avnishn March 16, 2022, 1:57am 2

It seems that this question was solved in this thread:

Topic		Replies	Views
Evaluation worker won't stop RLlib	3	611	June 19, 2022
Recommended way to evaluate training results RLlib	0	3329	June 12, 2021
Cannot create R2D2 trainer with evaluation worker RLlib	6	707	March 8, 2022
RLlib DQN Trainer Evaluate Function Help RLlib	1	354	August 22, 2022
Trainer.evaluate() runs 1 extra episode instead of as defined in evaluation_duration RLlib	1	393	August 26, 2022

DQNTrainer evaluate() doesn't perform any episode

Related topics