[Question:] How can we visualize trained agent? about elf HOT 5 CLOSED

@LinZichuan Currently you can specify --save_replay_prefix [replay_file_prefix] to save replays from your training/evaluation/selfplay. Then when you run backend, load the replay with the following command to visualize.

./minirts-backend replay --load_replay [your replay] --vis_after 0

from elf.

@yuandong-tian Thanks. After I add --save_replay_prefix, there are many replay output files. Should I use all of them? like --load_replay prefix_* ?
By the way, ./minirts-backend selfplay/humanplay --vis_after 0 do not work normally on frontend visualization on master branch after the merge. There is nothing when I open minirts.html using Firefox.
I tested on Ubuntu 16.04.2, cmake 3.5.1, gcc 5.4.0.

from elf.

@LinZichuan you can use any of them. They are generated in parallel.

from elf.

@qucheng might be you can take a look

from elf.

Should be fixed by latest commit

from elf.