def main():
parser = argparse.ArgumentParser(
description=
'Evaluate ProgressiveGAN w/ Hessian Penalty on Various Metrics')
# Model/Dataset Parameters:
parser.add_argument(
'--model',
required=True,
help=
'Either the number of experiment in results directory or a path to a .pkl checkpoint.'
)
parser.add_argument('--num_gpus',
type=int,
required=True,
help='Number of GPUs to evaluate with')
parser.add_argument(
'--snapshot_kimg',
default='latest',
help='network-snapshot-<snapshot_kimg>.pkl to evaluate')
parser.add_argument(
'--dataset',
type=str,
default='edges_and_shoes',
help=
'Name of TFRecords directory in datasets/ folder to run metrics with.')
parser.add_argument('--resolution',
type=int,
default=128,
help='Resolution of real data to evaluate with.')
parser.add_argument('--metrics',
nargs='+',
default=['FID', 'PPL'],
help='Metrics to run. Must specify at least one')
opt = parser.parse_args()
opt = EasyDict(vars(opt))
assert opt.num_gpus in [1, 2, 4, 8]
if os.path.isdir(
opt.model
): # If you pass a directory organized as DIRECTORY/dataset/model.pkl, it will iterate:git
paths = sorted(glob(f'{opt.model}/*/*.pkl'))
for path in paths:
opt.model = path
if 'clevr_simple' in path or 'clevr_u' in path:
opt.dataset = 'clevr_simple'
elif 'clevr_complex' in path:
opt.dataset = 'clevr_two_obj'
elif 'edgeshoes' in path:
opt.dataset = 'edges_and_shoes'
elif 'clevr_1fov' in path:
opt.dataset = 'clevr_1fov'
else:
print(f'Couldn\'t find dataset for {path}')
raise NotImplementedError
run(opt)
else:
run(opt)
