File size: 2,503 Bytes
462abe2 |
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 |
import os
import random
import shutil
import argparse
from pathlib import Path
from tqdm import tqdm
def SplitDataset(dataset_name: str, dataset_dir: str, val_ratio: float, seed: int = 42):
# Validate ratio
if not 0 < val_ratio < 1:
raise ValueError("Validation split ratio (--split) must be between 0 and 1 (exclusive).")
# Set paths
dataset_path = Path(dataset_dir) / dataset_name
train_path = dataset_path / "Train"
val_path = dataset_path / "Val"
# Validate dataset directory
if not dataset_path.exists():
raise FileNotFoundError(f"Dataset folder '{dataset_path}' not found.")
# Create Train/Val subfolders
train_path.mkdir(exist_ok=True)
val_path.mkdir(exist_ok=True)
# Find .parquet files
parquet_files = [f for f in dataset_path.glob("*.parquet")]
if not parquet_files:
raise ValueError(f"No .parquet files found in '{dataset_path}'.")
# Shuffle files deterministically
random.seed(seed)
random.shuffle(parquet_files)
# Split into train and val
val_count = int(len(parquet_files) * val_ratio)
val_files = parquet_files[:val_count]
train_files = parquet_files[val_count:]
print(f"📦 Found {len(parquet_files)} parquet files in '{dataset_name}'")
print(f"🔀 Splitting {val_count} to 'Val' ({val_ratio:.0%}) and {len(train_files)} to 'Train'")
# Move files to Val
for file in tqdm(val_files, desc="📁 Moving to Val", unit="file"):
shutil.move(str(file), str(val_path / file.name))
# Move files to Train
for file in tqdm(train_files, desc="📁 Moving to Train", unit="file"):
shutil.move(str(file), str(train_path / file.name))
print("✅ Split complete.")
if __name__ == "__main__":
parser = argparse.ArgumentParser(description="Split a dataset of .parquet files into Train and Val sets.")
parser.add_argument(
"-n", "--dataset_name", required=True, type=str,
help="Name of the dataset folder inside the dataset directory."
)
parser.add_argument(
"-d", "--dataset_dir", default="./src/Datasets", type=str,
help="Path to the parent directory containing dataset folders. Default is './src/Datasets'."
)
parser.add_argument(
"-s", "--split", default=0.1, type=float,
help="Fraction of files to use for validation (between 0 and 1). Default is 0.1 (10%)."
)
args = parser.parse_args()
SplitDataset(args.dataset_name, args.dataset_dir, args.split)
|