question

Keter asked Jul 4, '24 Keter commented Jul 5, '24

Reinforcement learning python observation expected

Hello,

I encountered some problems while initiating flexsim_training.py. I read the article that has similar problem. I know the issue is about the version of package and python. I already changed my python version to 3.7.8, stable-baselines3 to 2.0.0, and gymnasium to 0.28.0. However, I still get the same error. Hope I can get some advice, thanks!

The observation returned.PNG

flexsim_training.py

Software Version:

FlexSim 23.1.4

reinforcement learning python stable-baselines3 gym

the-observation-returned.png (37.8 KiB)

piplist.png (31.9 KiB)

flexsim-env.py (7.5 KiB)

flexsim-training.py (1.6 KiB)

rl-first.fsm (28.8 KiB)

Felix Möhlmann answered Jul 5, '24 Keter commented Jul 5, '24

Did the presented solution of specifying the dtype not work?

See also these two more recent posts:
https://answers.flexsim.com/questions/167193/reinforcement-learning-error-by-more-decision-even.html

https://answers.flexsim.com/questions/167263/reinforcement-learning-problem-by-test-runs.html

· 7

Keter commented · Jul 05 at 06:56 AM

Hi Mr. Möhlmann,

Can you provide your package version list to me? Cause I'm using gymnasium instead of gym.

0 ·

Felix Möhlmann Keter commented · Jul 05 at 07:15 AM

I can but it won't help much. I haven't really worked with the RL since that post and still have the old "gym" package.

The user in the other two posts I linked does use gymnasium though and got it work by setting the dtype. So that's why I asked if you had tried this.

0 ·

pip-list.png (34.9 KiB)

Keter Felix Möhlmann commented · Jul 05 at 08:09 AM

Yes, I tried set the dtype in the env.pydtype.PNG

0 ·

dtype.png (166.4 KiB)

Show more comments