- User Since
- Aug 1 2016, 12:10 PM (209 w, 1 d)
Nov 14 2019
Sep 27 2019
Option for browser rewrite T25582
Do this on first startup of browser/implement module check in browser?
Do this for browser rewrite T25582
Jul 12 2019
Jul 11 2019
Apr 30 2019
Apr 26 2019
No longer relevant
There is a setup option for .run_test, just set that to False
Mar 26 2019
Jan 24 2019
Does PyTorch offer a way to serialize DataLoader objects? If so, we could extend ModuleMultiTypeEncoder/Decoder in util.util
Jan 18 2019
Already works :D
Jan 17 2019
Oct 2 2018
Sep 14 2018
Aug 29 2018
Please provide more info. Do you mean a utility function that clears all environments? Why do we need that?
@zimmerer please close if this is resolved to your satisfaction
Current solution works alright for now, closing
Aug 28 2018
Jun 13 2018
Apr 24 2018
Apr 21 2018
Apr 20 2018
Apr 11 2018
I did two things that apparently solved the problem (>50k episodes without issue). 1. Move matplotlib.use("agg") to the __init__.py, because I was actually using Qt5Agg. This is a necessary fix regardless of this particular problem. 2. I "un-threaded" the saving of figures. This is likely the fix, but I still don't know why... Will mark as resolved regardless.
Apr 8 2018
Idea: Currently we're making .savefig threaded, maybe creating a little helper that does .savefig and then .clear in a thread could help?
Running python -m pdb myscript.py now gave out the following RuntimeWarning:
I can confirm that the segfault is in fact the problem. Running without visdom logging still results in segfault, but much later (sample size 1, ca. 13000 episodes). The EOFError is then only the result of the dead sender. Random idea from the internet: Stack memory limit too low? Mine is currently at 8192MB (ulimit -s)
Apr 6 2018
Mar 27 2018
Simple README not sufficient, Jupyter notebook better alternative.
Jan 18 2018
More or less ready, waiting for new wiki