def makeBaselinePolicy(_):
return BaselinePolicy()
extended_dir = zoo_dir / "extended"
# Example:
# reward_wrapper_dir = zoo_dir / "reward_wrapper"
obs_small_always_dir = zoo_dir / "obs_small_always"
# Add the test agents you would like to evaluate to the test_agents dictionary { "model_name": "path/to/model"}
# You may delete the existing ones in the test_agents dictionary (they are just examples)
test_agents = {
"ppo": PPOPolicy(obs_small_always_dir / "ppo.zip"),
"ppo_sp": PPOPolicy(obs_small_always_dir / "ppo_sp.zip"),
"ga_sp": makeSlimePolicyLite(obs_small_always_dir / "ga_sp.json"),
}
if args.benchmark and not args.evaltest:
pretrained_agents = {
"baseline": BaselinePolicy(),
"ppo": PPOPolicy(pretrained_dir / "ppo.zip"),
"ppo_sp": PPOPolicy(pretrained_dir / "ppo_sp.zip"),
"ga_sp": makeSlimePolicyLite(pretrained_dir / "ga_sp.json"),
"random": RandomPolicy(),
}
extended_agents = {
"ppo_extended": PPOPolicy(extended_dir / "ppo.zip"),
"ppo_sp_extended": PPOPolicy(extended_dir / "ppo_sp.zip"),
"ga_sp_extended": makeSlimePolicyLite(extended_dir / "ga_sp.json"),
}
benchmarks = {
"pretrained": pretrained_agents,
"extended": extended_agents