seermer / hollowknight_rl Goto Github PK

Sorry to bother you, I am very interested in RL related training, and I wanna apply it to train bosses instead of agent of player. Could I have some suggestions from you? btw, I can speak Mandarin, so we can talk on WeChat.

After correct configuration, if the code still raises the following error (hornet_new branch)

File "xxx\lib\site-packages\pyautogui\__init__.py", line 174, in wrapper
    raise ImageNotFoundException  # Raise PyAutoGUI's ImageNotFoundException.
pyautogui.ImageNotFoundException

Then it's probably due to pyautogui's update: https://pyautogui.readthedocs.io/en/latest/screenshot.html

NOTE: As of version 0.9.41, if the locate functions can’t find the provided image, they’ll raise ImageNotFoundException instead of returning None.

When loading the the bestmodel.pt, which was provided in the url:https://drive.google.com/drive/folders/1EOlSS6pknAID04wfB2AmB2XZCnuEHWxH?usp=sharing
by the author, or loading the latestoptimizer.pt which was trained by myself, the evaluate.py raised RuntimeError:

1. Missing key(s) in state_dict: "resetable.0.mu_W", "resetable.0.sigma_W", "resetable.0.mu_bias", ...
2. size mismatch for noisy.0.mu_W: copying a param with shape torch.Size([320, 11520]) from checkpoint, the shape in current model is torch.Size([512, 8000])...
   I wonder how to set the right model and weight. Thanks a lot!

Hello, I encountered a problem that I didn’t see the process of training after commander screening “training started”. It just keep pressing “w” and not gonna enter the fight.