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

committing changes to start testing framework, removing shuffling of data

Browse Source
This commit is contained in:
Arec
2022-01-17 15:47:40 -08:00
parent d34fa5774d
commit 2c1dc6ca33
3 changed files with 54 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files

41
scratch/arec/intersimple/test_model.py
View File

@@ -1,7 +1,9 @@
from tqdm import tqdm
from copy import deepcopy
import stable_baselines3 as sb3
import intersim
ALL_OPTIONS =
ALL_OPTIONS = [(v,t) for v in [0,2,4,6,8] for t in [5, 10]] # option 0 is safe fallback
def load_model(model_path:str, method:str):
"""
@@ -17,7 +19,7 @@ def load_model(model_path:str, method:str):
model = None
is_heir = False
if method == 'expert':
pass
raise NotImplementedError
elif method == 'bc':
raise NotImplementedError
elif method == 'gail':
@@ -26,7 +28,7 @@ def load_model(model_path:str, method:str):
raise NotImplementedError
elif method == 'hgail':
is_heir = True
raise NotImplementedError
model = sb3.PPO.load(model_path)
elif method == 'hrail':
is_heir = True
raise NotImplementedError
@@ -41,12 +43,20 @@ def load_expert_states(roundabout, track):
roundabout (str): roundabout name
track (str): track id
Returns:
expert_states (torch.tensor): (nv, T, 5) expert states for track file
states (torch.tensor): (T+1, nv, 5) expert states for track file
actions (torch.tensor): (T, nv, 1) expert actions for track file
"""
pass
state_path = '../../../expert_data/%s/track%04i/joint_expert_states.pt'%(roundabout, track)] #FIXME when moving
action_path = '../../../expert_data/%s/track%04i/joint_expert_actions.pt'%(roundabout, track)] #FIXME when moving
states = torch.load(path)
actions = torch.load(path)
# nanify actions where vehicle's don't exist
import pdb
pdb.set_trace()
return states, actions
def test_model(
locations=[],
locations=[(0,0)],
model_name='gail_image_multiagent_nocollision',
env='NRasterizedRouteIncrementingAgent',
method='expert',
@@ -56,7 +66,7 @@ def test_model(
Test a particular model at different locations/tracks
Args:
locations (list of tuples): list of (roundabout, track) pairs
locations (list of tuples): list of (roundabout, track) integer pairs
model_name (str): name of model to test
env (str): environment class
method (str): method (expert, bc, gail, rail, hgail, hrail)
@@ -70,23 +80,26 @@ def test_model(
all_vehicle_infos = []
for i, location in tqdm(enumerate(locations)):
# load expert states
expert_states = load_expert_states(roundabout, track)
# add roundabout and track to environent
roundabout, track = location
iround = intersim.LOCATIONS.index(roundabout)
it_env_kwargs = deepcopy(env_kwargs)
it_env_kwargs.update({})
loc_kwargs = {
'loc':iround,
'track':track
}
it_env_kwargs.update(loc_kwargs)
# load expert states and get average velocities
expert_states = load_expert_states(roundabout, track)
expert_states, expert_actions = load_expert_states(roundabout, track)
expert_vavg = torch.nanmean(expert_states[:,:,3], dim=-1)
# initialize environment
if not is_heir:
pass
Env = src.options.envs.__dict__[env]
else:
pass
Env = intersim.envs.intersimple.__dict__[env]
env = Env(**env_kwargs)
s = env.reset()
# Iterate through every vehicle and time

2
src/data/__init__.py
View File

@@ -1 +1 @@
from src.data.expert import single_agent_demonstrations, multi_agent_demonstrations, NoShuffleRNG, load_experts, process_experts
from src.data.expert import single_agent_expert, single_agent_demonstrations, multi_agent_demonstrations, NoShuffleRNG, load_experts, process_experts

39
src/data/expert.py
View File

@@ -112,7 +112,7 @@ class NoShuffleRNG(np.random.RandomState):
def shuffle(self, x):
return x
def load_experts(expert_files, flatten = True):
def load_experts(expert_files, flatten=True):
"""
Load expert trajectories from files and combine their transitions into a single RB
@@ -131,8 +131,27 @@ def load_experts(expert_files, flatten = True):
transitions = rollout.flatten_trajectories(transitions)
return transitions
def single_agent_demonstrations(expert='NormalizedIntersimpleExpert',
env='NRasterizedRouteIncrementingAgent',
def single_agent_expert(expert='NormalizedIntersimpleExpert',
env='NRasterizedRouteIncrementingAgent',
env_args={}, policy_args={}, **kwargs):
"""
Args:
expert (class): class of expert
env (class): class of env intersim.envs.intersimple
env_args (dict): dictionary of kwargs when instantiating environment class
policy_args (dict): dictionary of kwargs when instantiating Expert policy
path (str): path to store output
min_timesteps (int): min number of timesteps for call to rollout.rollout_and_save
min_episodes (int): min number of episodes for call to rollout.rollout_and_save
video (bool): whether to save a video of the expert until a single environment instantiation stops
"""
Env = intersim.envs.intersimple.__dict__[env]
Expert = globals()[expert]
env = Env(**env_args)
policy = Expert(env, **policy_args)
single_agent_demonstrations(env, policy, **kwargs)
def single_agent_demonstrations(env, policy,
path=None, min_timesteps=None,
min_episodes=None, video=False,
env_args={}, policy_args={}):
@@ -141,8 +160,8 @@ def single_agent_demonstrations(expert='NormalizedIntersimpleExpert',
Usage:
python -m intersimple.expert <flags>
Args:
expert (class): class of expert
env (class): class of env intersim.envs.intersimple
env (class): intersimple environment
policy (BasePolicy): intersimple policy
path (str): path to store output
min_timesteps (int): min number of timesteps for call to rollout.rollout_and_save
min_episodes (int): min number of episodes for call to rollout.rollout_and_save
@@ -151,14 +170,8 @@ def single_agent_demonstrations(expert='NormalizedIntersimpleExpert',
policy_args (dict): dictionary of kwargs when instantiating Expert policy
"""
Env = intersim.envs.intersimple.__dict__[env]
Expert = globals()[expert]
env = Env(**env_args)
info_env = RolloutInfoWrapper(env) # getting rollout info (dictionary) from environment
venv = DummyVecEnv([lambda: info_env]) # making a DummyVecEnv with a list of a function that when called returns the rollout info
policy = Expert(env, **policy_args) # instantiate an expert policy from specified class with instantiated environment and policy kwargs
venv_policy = DummyVecEnvPolicy([lambda: policy]) # make a DummyVecEnvPolicy with a list of a function that when called returns the Expert policy
if min_timesteps is None and min_episodes is None:
@@ -166,7 +179,7 @@ def single_agent_demonstrations(expert='NormalizedIntersimpleExpert',
if video:
save_video(env, policy)
path = path or (policy.__class__.__name__ + '_' + env.__class__.__name__ + '.pkl')
suntil = rollout.make_sample_until(
min_timesteps=min_timesteps,
@@ -253,7 +266,7 @@ def process_experts(filename:str='expert.pkl',
policy_args=expert_args
)
# Single-Agent POV Demonstrations
single_agent_demonstrations(
single_agent_expert(
expert=expert_class,
env=env_class,
path=it_path,