RL_Procs and Img_srv

[Williokeef]

Hello Smellslikeml and Kindbot team
I'd like to start off with thanks for releasing this github repo. It's a real bummer this did not get the proper attention during your kickstart campaign. I would have definitely jumped on this product.

Please bare with me as I'm a bit of a novice on the programming totem pole and even less experience with ML/RL/DL.

I've finally installed all the required dependencies for kindbot. At the moment I'm hitting a brick wall with two issues.

One is when running the RL_procs, the TL optimistic-restore is halting the process with an error in which it cannot find the "my-policy.ckpt". This is understandable as I have no ckpt as this is the first run. How can I go about bypassing this? I've tried commenting out the restore in the name/main area with no luck as that portion of the script is checking if all the variable are in place for the restore and continuation of the Rl. (Hope my interpretation of the code is some what correct) lol
(Edited* When I do comment out line 120 optimistic restore it results in a list index out of range for obs = data_stream. Makes sense since their are no values for the kindbot_sql log as that is driven by the environ log from the Rl_procs rewards)

Another is when running img_srv.py. Darkflow is reporting an over-read error with the tiny-yolo model and weights I've downloaded from the DF repo. Ive tried all the methods from adding the expected read and found values into the loader.py self.offset as suggested by others in the DF repo. Still no luck.

I was really hoping for this to be as simple as described in the readme. Install...configure..and im up and running..lol 😆

This is all running on a RPi 3B with the latest buster.
If you need any specifics please let me know and I'll try to provide them.