File size: 2,247 Bytes
fbe8d49 | 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 | """Single-source baselines of the same size, for comparison with the mix.
python baselines.py 10M -> out/base_tinystories_10M, out/base_fineweb_edu_10M
Light cleaning only (ftfy + whitespace), no mix, no complexity or dedup filters:
this is what someone would get by just taking the dataset.
"""
import os
import sys
from pathlib import Path
ROOT = Path(__file__).resolve().parent.parent
_env = ROOT / ".env"
if _env.exists():
for _l in _env.read_text(encoding="utf-8").splitlines():
if "=" in _l and not _l.lstrip().startswith("#"):
_k, _v = _l.split("=", 1)
os.environ.setdefault(_k.strip(), _v.strip().strip("'\""))
import ftfy
from datasets import load_dataset
from huggingface_hub import HfFileSystem
EOT = "<|endoftext|>"
def tinystories():
buf = []
with HfFileSystem().open("datasets/roneneldan/TinyStories/TinyStoriesV2-GPT4-train.txt", "r",
encoding="utf-8", errors="replace") as f:
for line in f:
if EOT in line:
if buf:
yield " ".join(buf)
buf = []
else:
buf.append(line.strip())
def fineweb_edu():
for r in load_dataset("HuggingFaceFW/fineweb-edu", "sample-10BT", split="train", streaming=True):
if r["int_score"] >= 4:
yield r["text"]
SOURCES = {"tinystories": tinystories, "fineweb_edu": fineweb_edu}
def main():
size = sys.argv[1]
target = {"10M": 10e6, "100M": 100e6}[size]
for name, gen in SOURCES.items():
if len(sys.argv) > 2 and name not in sys.argv[2:]:
continue
out = ROOT / "out" / f"base_{name}_{size}"
out.mkdir(parents=True, exist_ok=True)
words, n = 0, 0
with (out / "train.txt").open("w", encoding="utf-8") as f:
for text in gen():
text = ftfy.fix_text(text).strip()
if not text:
continue
f.write(text + f"\n{EOT}\n")
words += len(text.split())
n += 1
if words >= target:
break
print(f"{name} {size}: {words/1e6:.2f}M words, {n} docs", flush=True)
if __name__ == "__main__":
main()
|