batch = pa.RecordBatch.from_pandas(df)