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

updating bc experiment, adding automatic config copying, fixing idm/expert running from file issues in eval_experiments

Browse Source
This commit is contained in:
Arec Jamgochian
2022-02-28 18:56:28 -08:00
parent e602aa0641
commit 5f6ad37c37
4 changed files with 19 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
bc-experiment.py
View File

@@ -142,7 +142,7 @@ if __name__ == '__main__':
import argparse import argparse
parser = argparse.ArgumentParser() parser = argparse.ArgumentParser()
parser.add_argument('--train', choices=['A', 'B']) parser.add_argument('--train', choices=['A', 'B'])
parser.add_argument('--epochs', type=int, default=1000) parser.add_argument('--epochs', type=int, default=500)
parser.add_argument('--test', type=str, help='path to config file to run final training on') parser.add_argument('--test', type=str, help='path to config file to run final training on')
parser.add_argument('--test_seeds', type=int, default=5) parser.add_argument('--test_seeds', type=int, default=5)
parser.add_argument('--test_cpus', type=int, help='number of cpus available to split test seed training over') parser.add_argument('--test_cpus', type=int, help='number of cpus available to split test seed training over')
@@ -163,9 +163,9 @@ if __name__ == '__main__':
'policy': { 'policy': {
'learning_rate': 3e-4, 'learning_rate': 3e-4,
'learning_rate_decay': tune.grid_search([0.999, 1.0]), 'learning_rate_decay': tune.grid_search([0.999, 1.0]),
'hidden_layer_size': tune.grid_search([10, 20, 40, 80]), 'hidden_layer_size': tune.grid_search([10, 20, 40]),
'n_hidden_layers': tune.grid_search([2, 3, 4]), 'n_hidden_layers': tune.grid_search([2, 3]),
'activation':0, 'activation':tune.grid_search([0, 1]),
}, },
'train_epochs': args.epochs, 'train_epochs': args.epochs,
'seed': 0, 'seed': 0,
@@ -210,3 +210,5 @@ if __name__ == '__main__':
check_dir = analysis._checkpoints[i]['logdir'] check_dir = analysis._checkpoints[i]['logdir']
shutil.copyfile(os.path.join(check_dir,'policy_final.pt'), shutil.copyfile(os.path.join(check_dir,'policy_final.pt'),
os.path.join(savepath, f'policy_seed{s}.pt')) os.path.join(savepath, f'policy_seed{s}.pt'))
shutil.copyfile(os.path.join(check_dir,'params.json'),
os.path.join(savepath, 'config.json')) # copy config automatically

14
eval_experiments.py
View File

@@ -35,12 +35,12 @@ def main(method:str='expert', folder:str=None, locations=[(0,0)], skip_running=F
print('%i policy files found in %s folder' %(len(files), folder)) print('%i policy files found in %s folder' %(len(files), folder))
print('found policy config', config['policy']) print('found policy config', config['policy'])
policy_config = {k: v for k, v in config['policy'].items() if k not in exclude_keys_from_policy_kwargs} policy_config = {k: v for k, v in config['policy'].items() if k not in exclude_keys_from_policy_kwargs}
policy_config['activation'] = activations[policy_config['activation']] policy_config['activation'] = activations[policy_config['activation']]
print('final policy config', policy_config) print('final policy config', policy_config)
policy_kwargs.update(policy_config) policy_kwargs.update(policy_config)
print('final policy kwargs', policy_kwargs) print('final policy kwargs', policy_kwargs)
if not skip_running: if not skip_running:
for policy_file in files: for policy_file in files:
@@ -77,7 +77,7 @@ def latex_print(am, light=False):
print('success rate, distance travelled, RWSE_10, |DeltaV|, AccelJSD') print('success rate, distance travelled, RWSE_10, |DeltaV|, AccelJSD')
if light: if light:
if 'rwse_10s' in am.keys(): if 'rwse_10s' in am.keys():
print("%2.1f& %2.1f & %1.2f & %2.1f& " print("%2.1f& %2.1f & %2.1f & %1.2f& "
"%0.3f \\\\" %( 100*am['success rate'][0], am['mean travel distance'][0], am['rwse_10s'][0], "%0.3f \\\\" %( 100*am['success rate'][0], am['mean travel distance'][0], am['rwse_10s'][0],
am['average absolute average velocity'][0],am['acceleration distribution divergence'][0] )) am['average absolute average velocity'][0],am['acceleration distribution divergence'][0] ))
return return
@@ -88,7 +88,7 @@ def latex_print(am, light=False):
return return
print("%2.1f \\scriptstyle\\pm %2.1f & %2.1f \\scriptstyle\\pm %2.1f & " print("%2.1f \\scriptstyle\\pm %2.1f & %2.1f \\scriptstyle\\pm %2.1f & "
"%1.2f \\scriptstyle\\pm %1.2f & %2.1f \\scriptstyle\\pm %1.1f & " "%2.1f \\scriptstyle\\pm %1.1f & %1.2f \\scriptstyle\\pm %1.2f & "
"%0.3f \\scriptstyle\\pm %0.3f \\\\" %( 100*am['success rate'][0], 100*am['success rate'][1], "%0.3f \\scriptstyle\\pm %0.3f \\\\" %( 100*am['success rate'][0], 100*am['success rate'][1],
am['mean travel distance'][0] , am['mean travel distance'][1] , am['mean travel distance'][0] , am['mean travel distance'][1] ,
am['rwse_10s'][0] , am['rwse_10s'][1] , am['rwse_10s'][0] , am['rwse_10s'][1] ,

2
gail-experiment.py
View File

@@ -236,3 +236,5 @@ if __name__ == '__main__':
check_dir = analysis._checkpoints[i]['logdir'] check_dir = analysis._checkpoints[i]['logdir']
shutil.copyfile(os.path.join(check_dir,'policy_final.pt'), shutil.copyfile(os.path.join(check_dir,'policy_final.pt'),
os.path.join(savepath, f'policy_seed{s}.pt')) os.path.join(savepath, f'policy_seed{s}.pt'))
shutil.copyfile(os.path.join(check_dir,'params.json'),
os.path.join(savepath, 'config.json')) # copy config automatically

2
shail-experiment.py
View File

@@ -259,3 +259,5 @@ if __name__ == '__main__':
check_dir = analysis._checkpoints[i]['logdir'] check_dir = analysis._checkpoints[i]['logdir']
shutil.copyfile(os.path.join(check_dir,'policy_final.pt'), shutil.copyfile(os.path.join(check_dir,'policy_final.pt'),
os.path.join(savepath, f'policy_seed{s}.pt')) os.path.join(savepath, f'policy_seed{s}.pt'))
shutil.copyfile(os.path.join(check_dir,'params.json'),
os.path.join(savepath, 'config.json')) # copy config automatically