@@brthor1117 I quickly edited my comment after posting to remove my question. It's weird that you can still see it. At 16:26 you were working on the line. I ended up updating my epsilon in my act method because it seemed more logical that way. You probably just left it in the main training loop because it was still there from the previous video.
@@moikli I can only see the edit now 🤷♂️. For the epsilon, if you don't multiply the step number by the number of parallel environments, it will decay more slowly with respect to the number of actual environment steps we are taking. It should still work but your agent will improve more slowly. Since I introduced parallel environments in this video it was necessary to make the change.
There is something I do not understand. The Nature paper says that it stacks the most recent 4 frames. But what you are doing is to play 4 different games in parallel. So it is not quite the same with the Nature paper, is it?
hello sir, i am currently trying to learn to train dqn learning on atari using ubuntu but i keep getting the error "BrokenPipeError: [errno 32] Broken pipe". what would you suggest i do to fix this?
Windows user might get an error : FileNotFoundError: Could not find module 'C:\Users\"Your username"\PycharmProjects\"Your Project"\venv\lib\site-packages\atari_py\ale_interface\ale_c.dll' (or one of its dependencies). Just follow the steps here, stackoverflow.com/questions/63080326/could-not-find-module-atari-py-ale-interface-ale-c-dll-or-one-of-its-dependenc put the ale_c.dll in your project file will solve the erro
![[Classic] Playing Atari with Deep Reinforcement Learning (Paper Explained)](http://i.ytimg.com/vi/rFwQDDbYTm4/mqdefault.jpg)
![Reinforcement Learning Deep Q-Learning PyTorch Tutorial [Part 4]: Double Q-Learning (DDQN)](/img/n.gif)
Thanks for making this! I knew nothing about tensors and stuff but watching your videos on repeat has really helped me understand.
Glad you found it useful! Can you point to a time in the video you are referring to with the epsilon?
@@brthor1117 I quickly edited my comment after posting to remove my question. It's weird that you can still see it. At 16:26 you were working on the line. I ended up updating my epsilon in my act method because it seemed more logical that way. You probably just left it in the main training loop because it was still there from the previous video.
@@moikli I can only see the edit now 🤷♂️. For the epsilon, if you don't multiply the step number by the number of parallel environments, it will decay more slowly with respect to the number of actual environment steps we are taking. It should still work but your agent will improve more slowly. Since I introduced parallel environments in this video it was necessary to make the change.
thank you very much for your video, do you have the final code to download?
There is something I do not understand.
The Nature paper says that it stacks the most recent 4 frames.
But what you are doing is to play 4 different games in parallel.
So it is not quite the same with the Nature paper, is it?
Thanks for making the video. How can i display the GUI window when i run the observe.py code. (I have finished the trainning)
I found this very helpful, thanks for making this!
Glad it was helpful!
15:15 bro why did you bother updating the if statement earlier if you just going to delete it, had me so confused 😭
I dont get the video with agent playing the game when I run the observe file! can you help me with that @brthor
amazing! thanks from brazil!
Very cool! You're welcome from USA.
hello sir, i am currently trying to learn to train dqn learning on atari using ubuntu but i keep getting the error "BrokenPipeError: [errno 32] Broken pipe". what would you suggest i do to fix this?
From where I can get the code?
check the next video in the series
A video on rainbow ensemble please
gpu part starts from 36:43
Windows user might get an error :
FileNotFoundError: Could not find module 'C:\Users\"Your username"\PycharmProjects\"Your Project"\venv\lib\site-packages\atari_py\ale_interface\ale_c.dll' (or one of its dependencies).
Just follow the steps here,
stackoverflow.com/questions/63080326/could-not-find-module-atari-py-ale-interface-ale-c-dll-or-one-of-its-dependenc
put the ale_c.dll in your project file will solve the erro
I highly recommend linux for ai development in general!