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

getting loading of configs, overwriting with seeds, running with tune, and moving back to directory working. adding check for either training or testing, and allowing specification of number of test cpus to split seeds over

Browse Source
This commit is contained in:
Arec Jamgochian
2022-02-27 02:25:16 -08:00
parent 21cbd1c956
commit 569e0756ca
21 changed files with 27 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

39
shail-experiment.py
View File

@@ -142,8 +142,8 @@ def training_function(config):
v_opt=v_opt,
v_iters=config['value']['iterations_per_epoch'],
epochs=config['train_epochs'],
rollout_episodes=6, #60, FIXME
rollout_steps=6, #60, FIXME
rollout_episodes=60,
rollout_steps=60,
gamma=0.99,
gae_lambda=0.9,
clip_ratio=config['policy']['clip_ratio'],
@@ -160,19 +160,22 @@ def training_function(config):
if __name__ == '__main__':
import argparse
parser = argparse.ArgumentParser()
parser.add_argument('experiment', choices=['A', 'B'])
parser.add_argument('--train', choices=['A', 'B'])
parser.add_argument('--epochs', type=int, default=200)
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_cpus', type=int, help='number of cpus available to split test seed training over')
args = parser.parse_args()
assert (args.train is None) ^ (args.test is None), 'Must either train on an experiment or test with a config file'
# if no test config specified, train
if args.test is None:
print('Running Tuning for Experiment %s'%(args.experiment))
print('Running Tuning for Experiment %s'%(args.train))
analysis = tune.run(
training_function,
config={
'experiment': args.experiment,
'experiment': args.train,
'trainenv': {
'stop_on_collision': False,
'safe_actions_collision_method': 'circle',
@@ -213,13 +216,13 @@ if __name__ == '__main__':
os.mkdir(os.path.join(DIR, 'best_configs'))
# save shail
with open(os.path.join(DIR, 'best_configs',f'shail_exp{args.experiment}.json'), 'w', encoding='utf-8') as f:
with open(os.path.join(DIR, 'best_configs',f'shail_exp{args.train}.json'), 'w', encoding='utf-8') as f:
json.dump(best_config, f, ensure_ascii=False, indent=4)
# save hail
best_config['trainenv']['safe_actions_collision_method']=None
best_config['trainenv']['abort_unsafe_collision_method']=None
with open(os.path.join(DIR, 'best_configs',f'hail_exp{args.experiment}.json'), 'w', encoding='utf-8') as f:
with open(os.path.join(DIR, 'best_configs',f'hail_exp{args.train}.json'), 'w', encoding='utf-8') as f:
json.dump(best_config, f, ensure_ascii=False, indent=4)
# if config file specified, rerun it with appropriate number of seeds
@@ -227,13 +230,25 @@ if __name__ == '__main__':
with open(args.test, 'rb') as f:
config = json.load(f)
print(f'Retraining {args.test} with {args.test_seed} seeds on experiment {config["experiment"]}')
print(f'Retraining {args.test} with {args.test_seeds} seeds on experiment {config["experiment"]}')
# rerun with appropriate number of seeds
rpt = {'cpu': int(args.test_cpus/args.test_seeds)} if (args.test_cpus is not None) else None
config['seed'] = tune.grid_search(list(range(1,args.test_seeds+1)))
analysis = tune.run(training_function, config)
analysis = tune.run(training_function, config=config, resources_per_trial=rpt)
# move final policies to appropriate directory
# import pdb
# pdb.set_trace()
# a = 0
split_ = os.path.basename(args.test).split('_')
model = split_[0]
exper = split_[-1].split('.')[0]
savepath = os.path.join('test_policies',model,exper)
if not os.path.isdir(savepath):
os.makedirs(savepath)
import shutil
for i in range(args.test_seeds):
s = analysis._checkpoints[i]['config']['seed']
check_dir = analysis._checkpoints[i]['logdir']
shutil.copyfile(os.path.join(check_dir,'policy_final.pt'),
os.path.join(savepath, f'policy_seed{s}.pt'))