search

rlworkgroup / benchmarks

Continuous benchmarking of popular reinforcement learning algorithms
2 stars 0 forks source link

Download trained agents #6

Open ryanjulian opened 5 years ago

cheng-kevin commented 5 years ago

Is this the same as the policy files?

cheng-kevin commented 5 years ago

I assume we only want to make available the trained garage agents? and not the baselines/rlkit/etc ?

ryanjulian commented 5 years ago

if baselines/rlkit provide a decent interface for doing this, i think yes we should provide them.

the exact form of the file depends on the algo library. for garage, it is the .pkl files produced by training.