huangfu/InteractionImitation
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge branch 'main' into idm_upgrade

Browse Source
This commit is contained in:
Johannes Fischer
2022-03-01 09:58:07 +01:00
parent 2965dc9982 57a42f70ec
commit f93e130498
41 changed files with 94 additions and 298 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
src/eval_main.py
View File

@@ -41,33 +41,33 @@ def load_policy(method:str,
if method == 'idm':
policy = IDMRulePolicy(env, **policy_kwargs)
elif method == 'bc':
policy = SetPolicy(env.action_space.shape[-1])
policy = SetPolicy(env.action_space.shape[-1], **policy_kwargs)
policy.load_state_dict(torch.load(policy_file, map_location=ml))
policy.eval()
elif method == 'gail-trpo':
policy = SetPolicy(env.action_space.shape[-1])
policy = SetPolicy(env.action_space.shape[-1], **policy_kwargs)
policy(torch.zeros(env.observation_space.shape))
policy = ReparamPolicy(policy)
policy.load_state_dict(torch.load(policy_file, map_location=ml))
policy.eval()
elif method == 'gail':
policy = SetPolicy(env.action_space.shape[-1])
policy = SetPolicy(env.action_space.shape[-1], **policy_kwargs)
policy.load_state_dict(torch.load(policy_file, map_location=ml))
policy.eval()
elif method == 'rail':
raise NotImplementedError
elif method == 'hail-trpo':
policy = SetDiscretePolicy(env.action_space.n)
policy = SetMaskedDiscretePolicy(env.action_space.n, **policy_kwargs)
policy(torch.zeros(env.observation_space.shape))
policy = ReparamPolicy(policy)
policy.load_state_dict(torch.load(policy_file, map_location=ml))
policy.eval()
elif method == 'hail':
policy = SetDiscretePolicy(env.action_space.n)
policy = SetMaskedDiscretePolicy(env.action_space.n, **policy_kwargs)
policy.load_state_dict(torch.load(policy_file, map_location=ml))
policy.eval()
elif method == 'shail-trpo':
policy = SetMaskedDiscretePolicy(env.action_space.n)
policy = SetMaskedDiscretePolicy(env.action_space.n, **policy_kwargs)
policy(
torch.zeros(env.observation_space['observation'].shape),
torch.zeros(env.observation_space['safe_actions'].shape)
@@ -76,7 +76,7 @@ def load_policy(method:str,
policy.load_state_dict(torch.load(policy_file, map_location=ml))
policy.eval()
elif method == 'shail':
policy = SetMaskedDiscretePolicy(env.action_space.n)
policy = SetMaskedDiscretePolicy(env.action_space.n, **policy_kwargs)
policy.load_state_dict(torch.load(policy_file, map_location=ml))
policy.eval()
else: