3 ms·
This is awesome and I hope will allow more people to experiment with algorithms, instead of only re-applying OpenAI's baselines. Baselines are great, but are ve
by dimitry12 8y ago
This is awesome and I hope will allow more people to experiment with algorithms, instead of only re-applying OpenAI's baselines. Baselines are great, but are very hard (for me, at least) to tinker with.
It helps me to understand something new if I can controllably break it. In other words, I progress by predicting the edge-conditions when something shouldn't work - and then testing if algorithm indeed experienced expected type of failure. Transparent algorithm implementation is key for this.
One thing, which I immediately checked in the spinningup-repo is if it uses TF Eager. And it doesn't. @OpenAI what's your reasoning for that?
- jachiam 8y agoHi! Primary developer for Spinning Up here. The code for this was developed mostly in June and July this year, and Eager still felt relatively new to me. I wanted to wait for Eager to stabilize and hit maturity before investing in it. I also wanted to see how TF would change on the road to TF 2.0, since that could change the picture even more. At the six month review in 2019, we'll evaluate whether it makes sense to rewrite the implementation examples for TF 2.0. I'll speculate that the answer will be "yes, it does." Since Eager execution will be a central feature of TF 2.0, the (probable) revamp for Spinning Up will include it. Good luck with your experiments! And please let us know about your experience with Spinning Up---we want to make sure it fulfills the mission of helping anyone learn about deep RL, and user feedback is vital for that.
- dimitry12 8y agoThank you for sharing your thought process!