search

ZhengyaoJiang / PGPortfolio

PGPortfolio: Policy Gradient Portfolio, the source code of "A Deep Reinforcement Learning Framework for the Financial Portfolio Management Problem"(https://arxiv.org/pdf/1706.10059.pdf).
GNU General Public License v3.0
1.74k stars 750 forks source link

Mac OS X support #12

Closed srozov closed 6 years ago

srozov commented 6 years ago

First of all, I'd like to thank you for your work! I was able to run the code in Windows, the only minor issue was to install some modules for python 3.5, had to find some unofficial wheels. However, I couldn't make it run on MacOS X, it's crashing due to an unknown error (¨python quit unexpectedly¨). I know this is not officially supported, but I'd like to know if you have any hints on what might be the issue... Here's what I already tried: -Deleted .DS_Store files from all subdirectories (so that it passes the if not str.isdigit(dir): condition) -Tried different python versions and clean environments It fails to executeTraderTrainer(config, save_path=save_path, device=device).train_net(log_file_dir=log_file_dir, index=index). The log file is created but remains empty... Do you think this has something to do with pathname issues in OS X or the python modules?

srozov commented 6 years ago

Ok, so apparently it crashes after the line vol = self._polo.marketVolume() in coinlist.py. Probably has something to do with the poloniex wrapper. Edit: Got it to work. Tried with the release v1 and Data.db and setting "online":false after generating. If I set it before it shows Value error: the length of selected coins 0 is not equal to expected 11

ZhengyaoJiang commented 6 years ago

So, it's a good news that the program could work on Mac OS.