data/clip_metadata.py

import pandas as pd

_SPLIT = "train"

df = pd.read_parquet("data/dataset_audio_" + _SPLIT +".parquet.gzip")

clipped_df = df.filter(["label_string", "probability", "probability_vector", "prediction",
                "prediction_string", "embedding_reduced"], axis=1)

clipped_df = clipped_df.sort_values(by=["label_string"])
clipped_df.to_parquet("data/dataset_audio_" + _SPLIT +"_clipped.parquet.gzip")