I wonder if anybody gets better results than I do #32
Replies: 4 comments
-
Hi @tsadigovAgmail ! Chapter 6 is primarily focused on the Deep Q-Learner so, I am assuming you are running the deep_Q_learner.py script. By default, the agent will train in the If you would like to dig more into your observation, please share the parameters and configs you are using so that we can take a closer look. |
Beta Was this translation helpful? Give feedback.
-
Thanks @praveen-palanisamy How I came here is I wanted to make sure I interpret chart in the right way so I disabled learning part to have some baseline. I was wrongly interpreting it as result of epsilon decreasing and model stabilizing. |
Beta Was this translation helpful? Give feedback.
-
Do you have a sample chart of what should I see while the model really learns and how long can I expect it to take? |
Beta Was this translation helpful? Give feedback.
-
My current setup The baseline_NOT_learning shows fixed performance as expected (I did 4 runs with similar result not showing in screenshot) but I dont understand why all of the learning versions degrade in performance ? Do I need to just wait for longer? I would like to compare to results you had. |
Beta Was this translation helpful? Give feedback.
-
I ran the example for ch6 10 times,
Then changed config so that it does not learn, I mean does not update weights.
The results seem very similar. It averages between 35-40. I wonder if anybody gets better results than I do. May be I am applying example wrong way.
Beta Was this translation helpful? Give feedback.
All reactions