Datasets:
from pathlib import Path | |
import mteb | |
log_file_path = Path("remove_empty.log") | |
# remove log file if exists | |
if log_file_path.exists(): | |
log_file_path.unlink() | |
tasks = mteb.get_tasks(tasks=["STS22"]) | |
from datasets import load_dataset | |
dataset = load_dataset(**tasks[0].metadata.dataset) | |
def filter_sample(x): | |
if len(x["sentence1"]) > 0 and len(x["sentence2"]) > 0: | |
return True | |
log = f"Filtered: {x['sentence1']} -- {x['sentence2']}" | |
with open(log_file_path, "a") as f: | |
f.write(log + "\n") | |
print(log) | |
return False | |
for split in dataset: | |
ds = dataset[split] | |
# filter empty sentences | |
n_samples = len(ds) | |
ds = ds.filter(lambda x: filter_sample(x)) | |
n_left = len(ds) | |
log = f"Filtered {n_samples - n_left} samples from {n_samples} in {split}" | |
with open(log_file_path, "a") as f: | |
f.write(log + "\n") | |
print(log) | |
dataset[split] = ds | |
save_path = Path(__file__).parent.parent / "data" | |
for split in dataset: | |
# dataset[split].to_parquet(save_path / f"{split}-00000-of-00001.parquet") | |
dataset[split].to_json(save_path / f"{split}.jsonl.gz", compression="gzip") | |
ds = load_dataset(tasks[0].metadata.dataset["path"]) | |