elif param in float_params:
conf_dict[param] = float(c[param])
else:
conf_dict[param] = c[param]
return conf_dict
parser = argparse.ArgumentParser(description='Run training')
parser.add_argument("--config",
type=str,
default="./src/df_v1/config/config_heart.conf",
help="Path to the config file.")
parser.add_argument("--data.dataset", type=str, default=None)
parser.add_argument("--data.split", type=str, default=None)
parser.add_argument("--data.batch_size", type=int, default=None)
parser.add_argument("--data.episodes", type=int, default=None)
parser.add_argument("--data.cuda", type=int, default=None)
parser.add_argument("--data.gpu", type=int, default=None)
parser.add_argument("--train.patience", type=int, default=None)
parser.add_argument("--train.lr", type=float, default=None)
# Run training
args = vars(parser.parse_args())
config = configparser.ConfigParser()
config.read(args['config'])
filtered_args = dict((k, v) for (k, v) in args.items() if not v is None)
config = preprocess_config({**config['TRAIN'], **filtered_args})
train(config)
test_query,
'data.train_size':
train_size,
'data.test_size':
test_size,
'data.rotation_range':
rotation_range,
'data.width_shift_range':
width_shift_range,
'data.height_shift_range':
height_shift_range,
'data.horizontal_flip':
horizontal_flip,
'model.type': model_type,
'model.nb_layers':
nb_layers,
'model.nb_filters':
nb_filters,
'train.lr': lr
}
preprocessed_config = preprocess_config({
**config_from_file['TRAIN'],
**custom_params
})
train(preprocessed_config)
except:
print(
"Error. Probably memory :c"
)