SB3 doesn't seem to log success rate

[npnike10]

❓ Question

I am using DQN and PPO. DQN shows success rate for me during training and in tensorboard logs, but PPO doesn't.

I saw that to have success rate logged and displayed during training, one needs to add a dict with 'is_success' key as the value of info at the end of episodes. So I did that and DQN showed me success rate during data it displays when training as well as in the tensorboard logs. I did not wrap env with Monitor myself or use any callbacks etc., I just constructed my gym env which gives such an info and passed it to DQN.

Now with PPO, I don't get the success rate. I am not sure why. As something to try, since the Logger section in SB3 docs suggests wrapping env with a Monitor and giving a particular argument to ensure success rate is logged, I tried that but what happened is the generated monitor.csv does contain a column 'is_success' with true/false values but the data displayed during training or tensorboard logs don't seem to contain success rate. Here's the monitor line:

env = Monitor(env, "./ppo_sumo_tensorboard/", info_keywords=("is_success",))

I don't know if any other code would be helpful but I can happily provide any if it is.

Thank you very much for your time!

Checklist

• I have checked that there is no similar issue in the repo
• I have read the documentation
• If code there is, it is minimal and working
• If code there is, it is formatted using the markdown code blocks for both code and stack traces.

[npnike10]

I see there are similar issues. Sorry.