[Bug] uses intent column as labels for training and not tag column

[biswaroop1547]

currently the template uses intent column as labels for training and not tag column whereas intent column actually now model predictions for new charon's dataframe output format.

more info

[biswaroop1547]

63 this PR fixes it

[biswaroop1547]

To convert a previously used csv into the new charon format use this snippet by putting the whole csv file path in CSV_PATH variable:

import pandas as pd
import json
import tqdm

CSV_PATH = ""

def new_charon_format_conversion(CSV_PATH):
    df = pd.read_csv(CSV_PATH)
    df.drop(columns=['tag'], inplace = True)
    df['tag'] = df['intent'].values

    for i, row in tqdm.tqdm(df.iterrows(), total = len(df)):
        try:
            intent_pred = json.loads(row['prediction'])['name']
        except:
            intent_pred = None

        df.iloc[i, df.columns.get_loc('intent')] = intent_pred

    df.to_csv(CSV_PATH[:-4] + 'new.csv', index = False)
    print(f"{CSV_PATH} saved!")

new_charon_format_conversion(CSV_PATH)