sumanthd/IndicTrans-MultilingualTranslation
6
1from tqdm import tqdm2import sys3 4 5def remove_large_sentences(src_path, tgt_path):6 count = 07 new_src_lines = []8 new_tgt_lines = []9 src_num_lines = sum(1 for line in open(src_path, "r", encoding="utf-8"))10 tgt_num_lines = sum(1 for line in open(tgt_path, "r", encoding="utf-8"))11 assert src_num_lines == tgt_num_lines12 with open(src_path, encoding="utf-8") as f1, open(tgt_path, encoding="utf-8") as f2:13 for src_line, tgt_line in tqdm(zip(f1, f2), total=src_num_lines):14 src_tokens = src_line.strip().split(" ")15 tgt_tokens = tgt_line.strip().split(" ")16 if len(src_tokens) > 200 or len(tgt_tokens) > 200:17 count += 118 continue19 new_src_lines.append(src_line)20 new_tgt_lines.append(tgt_line)21 return count, new_src_lines, new_tgt_lines22 23 24def create_txt(outFile, lines, add_newline=False):25 outfile = open("{0}".format(outFile), "w", encoding="utf-8")26 for line in lines:27 if add_newline:28 outfile.write(line + "\n")29 else:30 outfile.write(line)31 outfile.close()32 33 34if __name__ == "__main__":35 36 src_path = sys.argv[1]37 tgt_path = sys.argv[2]38 new_src_path = sys.argv[3]39 new_tgt_path = sys.argv[4]40 41 count, new_src_lines, new_tgt_lines = remove_large_sentences(src_path, tgt_path)42 print(f'{count} lines removed due to seq_len > 200')43 create_txt(new_src_path, new_src_lines)44 create_txt(new_tgt_path, new_tgt_lines)45 