Issues encountered when processing big data

When using the process.sh script, I can process my test and validation datasets normally, but I am unable to process my training dataset without any error output. I then added a batch field in extract.py and changed the directory scanning to batch scanning instead of scanning all at once. After these modifications, I was able to get the expected output for the training data. The modified py file is as follows,we add the parameter "batch_size" and update ExtractFeaturesForDirsList mehtod:


#!/usr/bin/python

import itertools
import multiprocessing
import os
import sys
import shutil
import subprocess
from threading import Timer
import sys
from argparse import ArgumentParser
from subprocess import Popen, PIPE, STDOUT, call

# ......

def ExtractFeaturesForDirsList(args, dirs):
    tmp_dir = f"./tmp/feature_extractor{os.getpid()}/"
    if os.path.exists(tmp_dir):
        shutil.rmtree(tmp_dir, ignore_errors=True)
    os.makedirs(tmp_dir)
    try:
        for i in range(0, len(dirs), args.batch_size):  # 使用range和batch_size来分批处理
            batch_dirs = dirs[i:i + args.batch_size]
            p = multiprocessing.Pool(4)
            p.starmap(ParallelExtractDir, zip(itertools.repeat(args), itertools.repeat(tmp_dir), batch_dirs))
            output_files = os.listdir(tmp_dir)
            for f in output_files:
                os.system("cat %s/%s" % (tmp_dir, f))
                os.remove(os.path.join(tmp_dir, f))  # 删除处理过的文件，为下一批做准备
    finally:
        shutil.rmtree(tmp_dir, ignore_errors=True)

if __name__ == '__main__':
    parser = ArgumentParser()
    parser.add_argument("-maxlen", "--max_path_length", dest="max_path_length", required=False, default=8)
    parser.add_argument("-maxwidth", "--max_path_width", dest="max_path_width", required=False, default=2)
    parser.add_argument("-threads", "--num_threads", dest="num_threads", required=False, default=64)
    parser.add_argument("-j", "--jar", dest="jar", required=True)
    parser.add_argument("-dir", "--dir", dest="dir", required=False)
    parser.add_argument("-file", "--file", dest="file", required=False)
    parser.add_argument("-batch_size", "--batch_size", dest="batch_size", required=False, default=5, type=int)

    args = parser.parse_args()

    if args.file is not None:
        command = 'java -cp ' + args.jar + ' JavaExtractor.App --max_path_length ' + \
                  str(args.max_path_length) + ' --max_path_width ' + str(args.max_path_width) + ' --file ' + args.file
        os.system(command)
    elif args.dir is not None:
        subdirs = get_immediate_subdirectories(args.dir)
        to_extract = subdirs
        if len(subdirs) == 0:
            to_extract = [args.dir.rstrip('/')]
        ExtractFeaturesForDirsList(args, to_extract)

tech-srl / code2vec

Issues encountered when processing big data #181