File size: 2,365 Bytes
8c941a2 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 |
import random
import glob
import os
import json
import argparse
from tqdm import tqdm
import umi_parse
import cornell_movie_dialogue_parse
import eratoho_parse
import valhalla_parse
import discord_parse
def compile_raw():
print('-- Parsing Umineko Data --\n')
umi_parse.parse()
print('\n-- Parsing Cornell Movie Dialogue Data --\n')
cornell_movie_dialogue_parse.parse()
print('\n-- Parsing Eratoho Data --\n')
eratoho_parse.parse()
print('\n-- Parsing Valhalla Data --\n')
valhalla_parse.parse()
print('\n-- Parsing Discord Data --\n')
discord_parse.parse()
def compile_mtf_jax():
# compiles dataset into a single text file to be tokenized by the mtf jax repo
# get all the files with *.txt in ./data.
files = glob.glob('data/*/*.txt')
with open('output.txt', 'w', encoding='utf-8') as f:
for file in tqdm(files):
with open(file, 'r', encoding='utf-8') as f2:
f.write(f2.read().replace('\n\n', '\n'))
f.write('\n')
# remove all double newlines
lines = ''
with open('output.txt', 'r', encoding='utf-8') as f:
lines = f.read().replace('\n\n', '\n')
with open('output.txt', 'w', encoding='utf-8') as f:
f.write(lines)
def compile_gpt_neo():
# compile each file into a json lines file
files = glob.glob('data/*/*.txt')
# shuffle the files
random.shuffle(files)
with open('output.jsonl', 'w', encoding='utf-8') as f:
for file in tqdm(files):
with open(file, 'r', encoding='utf-8') as f2:
f.write(json.dumps({'text': f2.read().replace('\n\n', '\n')}))
f.write('\n')
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Process raw data')
parser.add_argument('-d', '--dont_compile', action='store_true', help='dont compile raw data', default=False)
parser.add_argument('-m', '--mtf_jax', action='store_true', help='compile raw data into a single text file to be tokenized by the mtf jax repo')
parser.add_argument('-g', '--gpt_neo', action='store_true', help='compile raw data into a single json lines file')
args = parser.parse_args()
if not args.dont_compile:
compile_raw()
if args.mtf_jax:
compile_mtf_jax()
if args.gpt_neo:
compile_gpt_neo()
print('Done!')
|