Retro analyze fail gracefully when no experiment data

kengz / SLM-Lab

Modular Deep Reinforcement Learning framework in PyTorch. Companion library of the book "Foundations of Deep Reinforcement Learning".

https://slm-lab.gitbook.io/slm-lab/

MIT License

1.25k stars 266 forks source link

Closed lgraesser closed 5 years ago

lgraesser commented 5 years ago

guard the info_prepath look up in retro_analyze_experiment and log an appropriate message if none is found. It is misleading for the program to crash when an experiment wasn't run but the session and trial analyses succeeded.