PaddlePaddle / PARL

A high-performance distributed training framework for Reinforcement Learning
https://parl.readthedocs.io/
Apache License 2.0
3.22k stars 817 forks source link

fix `run_evaluate_episodes` parameter #1033

Open DrRyanHuang opened 1 year ago

DrRyanHuang commented 1 year ago

Hi, thx for PARL's good work!

The second para of run_evaluate_episodes is eval_episodes: int, so we don't need env.

And it will raise 'CompatWrapper' object cannot be interpreted as an integer in the 55th line of examples/QuickStart/train.py.

By the way, I add the cli --eval.