search

Zeta36 / chess-alpha-zero

Chess reinforcement learning by AlphaGo Zero methods.
MIT License
2.13k stars 481 forks source link

Lot of errors in opt #85

Open CasellaJr opened 4 years ago

CasellaJr commented 4 years ago

Hi everyone. I have problems with the command python src/chess_zero/run.py opt

First of all, the procedure I followed is this:

  1. I downloaded a dataset of 2000 pgn from ficsgames.org, and I put in the play_data folder
  2. Then I ran python src/chess_zero/run.py sl 5 times
  3. Then I ran python src/chess_zero/run.py self and I wait for 800 self games
  4. I ran python src/chess_zero/run.py opt Here, I have some problems, because I receive this error: https://pastebin.com/Px7N0S3U Some other times, the errors are related to the memory, but I don't know why they appear rarely, and 99% of times the error is TypeError: 'NoneType' object is not iterable

Someone knows what can I do for solving? Thanks

Akababa commented 4 years ago

It appears line 19 is the cause: 2019-12-13 22:03:31.118125: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'nvcuda.dll'; dlerror: nvcuda.dll not found I assume you're using tensorflow-gpu, did you install CUDA? https://www.tensorflow.org/install/gpu Alternatively you can try to run it using CPU-only by installing tensorflow (not tensorflow-gpu)