import environments
import experiments
from experiments.plotting import plot_policy_map, plot_policy_map1, \
plot_policy_map_combined, plot_value_map, plot_value_map1, plot_value_map2
base_dir = 'output/Q'
#base_dir = 'test-heatmap'
output_dir = 'output/images/Q'
#output_dir = 'test-heatmap'
envs = {}
envs['4x4'] = environments.get_small_frozen_lake_environment()
envs['8x8'] = environments.get_medium_rewarding_frozen_lake_environment()
envs['15x15'] = environments.get_large_frozen_lake_environment()
envs['4x12'] = environments.get_windy_cliff_walking_environment()
title_regex = re.compile('v-(.*)\.txt')
env_strs = ['4x4', '8x8', '15x15', '4x12']
#env_strs = ['4x12']
for env_str in env_strs:
grid_files = glob.glob('{}/v-*{}*.txt'.format(base_dir, env_str))
print(grid_files)
env = envs[env_str]
for path in grid_files:
file = basename(path)
search_result = title_regex.search(basename(file))
logger.info("Creating MDPs")
logger.info("----------")
envs = [{
'env': environments.get_large_rewarding_frozen_lake_environment(),
'name': 'large_frozen_lake',
'readable_name': 'Frozen Lake (18x20)',
}, {
'env':
environments.get_large_rewarding_frozen_lake_slipper_environment(),
'name':
'large_frozen_lake',
'readable_name':
'Frozen Lake (18x20)',
}, {
'env': environments.get_windy_cliff_walking_environment(),
'name': 'cliff_walking_windy',
'readable_name': 'Cliff Walking (4x12)',
}, {
'env': environments.get_cliff_walking_environment(),
'name': 'cliff_walking',
'readable_name': 'Cliff Walking (4x12)',
}]
experiment_details = []
for env in envs:
env['env'].seed(seed)
logger.info('{}: State space: {}, Action space: {}'.format(
env['readable_name'], env['env'].unwrapped.nS,
env['env'].unwrapped.nA))
experiment_details.append(
'readable_name':
'Frozen Lake (8x8)',
}, {
'env':
environments.get_large_rewarding_frozen_lake_environment(
ENV_REWARDS['large_lake']['step_rew'],
ENV_REWARDS['large_lake']['hole_rew'],
ENV_REWARDS['large_lake']['goal_rew']),
'name':
'large_frozen_lake',
'readable_name':
'Frozen Lake (15x15)',
}, {
'env':
environments.get_windy_cliff_walking_environment(
ENV_REWARDS['cliff_walking']['step_rew'],
ENV_REWARDS['cliff_walking']['fall_rew'],
ENV_REWARDS['cliff_walking']['goal_rew']),
'name':
'cliff_walking',
'readable_name':
'Cliff Walking (4x12)',
}]
experiment_details = []
for env in envs:
env['env'].seed(seed)
logger.info('{}: State space: {}, Action space: {}'.format(
env['readable_name'], env['env'].unwrapped.nS,
env['env'].unwrapped.nA))
experiment_details.append(
experiments.ExperimentDetails(env['env'],