mirror of
https://github.com/JasonYANG170/CodeGeeX4.git
synced 2024-11-23 20:26:29 +00:00
21 lines
845 B
Python
21 lines
845 B
Python
|
import argparse
|
||
|
|
|
||
|
|
from utils.data import traverse
|
||
|
|
from utils.vector import save_vectors
|
||
|
|
|
||
|
|
|
||
|
|
def parse_arguments():
|
||
|
|
parser = argparse.ArgumentParser()
|
||
|
|
parser.add_argument('--workspace', type=str, help="directory of the workspace to be vectorized", default='.')
|
||
|
|
parser.add_argument('--lines_per_chunk', type=int, help="chunk lines when splitting", default=40)
|
||
|
|
parser.add_argument('--lines_overlap', type=int, help="chunk lines overlap when splitting", default=15)
|
||
|
|
parser.add_argument("--max_chars", type=int, help="maximum number of characters in a chunk", default=1500)
|
||
|
|
parser.add_argument('--output_path', type=str, help="path to save the vectors", default='vectors')
|
||
|
|
return parser.parse_args()
|
||
|
|
|
||
|
|
|
||
|
|
if __name__ == '__main__':
|
||
|
|
args = parse_arguments()
|
||
|
|
files = traverse(args.workspace)
|
||
|
|
save_vectors(files, args)
|