search

hsahovic / poke-env

A python interface for training Reinforcement Learning bots to battle on pokemon showdown
https://poke-env.readthedocs.io/
MIT License
297 stars 103 forks source link

Converting showdown targets into array indices #523

Closed caymansimpson closed 7 months ago

caymansimpson commented 7 months ago

Right now, it's very hard to specify what mon should be targeting what other mon in double, since showdown targets don't map cleanly to arrays. I had to do a workaround mapping function solution in a forked version of poke-env here: https://github.com/caymansimpson/poke-env/blob/5f09a6d814b294c0c6827131485101afa2f64668/src/poke_env/utils.py#L90

I think it would be great to have "/move hyperbeam 0" to target the mon in the 0th position of the opponent's team's tuple

hsahovic commented 7 months ago

Yeah, this is pretty bad in terms of usability. What do you think would be the clearest API? Imo, it would be ideal to abstract away positions, for instance by intelligently handling the move_target arg in BattleOrder. Does that make sense?

caymansimpson commented 7 months ago

I think that makes sense -- so people will just use the index in provided in poke-env, and we will handle everything else? If so, how do we handle self-targets? (when battle.opponent_active_pokemon is 0 or 1, mapping to 1 and 2, and self-targets are -1 or -2)

Maybe enums would be best?

caymansimpson commented 7 months ago

Combining with: https://github.com/hsahovic/poke-env/pull/529