def test(method, bits, random_top_layer, quantize_top_layer, results={}):
    print(prompt)
    start_time = time.time()
    if method == 'vanilla':
        cnt_tokens = test_vanilla(bits)
    elif method == 'eagle':
        cnt_tokens = test_eagle(bits,
            random_top_layer=random_top_layer,
            quantize_top_layer=quantize_top_layer
        )
    time_delta = time.time() - start_time
    speed = cnt_tokens / time_delta
    print('e2e speed:', time_delta, cnt_tokens, speed)
    results.update(dict(time_delta=time_delta, cnt_tokens=cnt_tokens, speed=speed))


if __name__ == '__main__':
    import os
    import argparse
    import pandas as pd
    import multiprocessing, threading
    from colorama import Fore, Back, Style
    parser = argparse.ArgumentParser(description='Pandas Fire ArgumentParser')
    parser.add_argument('--debug', action='store_true')
    args = parser.parse_args()
    manager = multiprocessing.Manager()
    df_params = pd.read_csv('params-awq.tsv', sep='\t', header=0)
    df_params = df_params.replace({float('nan'): None})
    df_results = []
    for params in df_params.to_dict(orient='records'):
        print(Fore.RED, Back.YELLOW, params, Style.RESET_ALL)
        try:
            params['results'] = manager.dict()
            os.environ["TOKENIZERS_PARALLELISM"] = "true"
            if args.debug:
                process = threading.Thread(target=test, kwargs=params)
            else:
                process = multiprocessing.Process(target=test, kwargs=params)
            process.start()
            process.join()
        except:
            if args.debug:
                pass
            else:
                process.terminate()
            break
        results = dict(params['results'])
        print(Fore.RED, Back.YELLOW, results, Style.RESET_ALL, end='\n\n')
        df_results.append(results)
    df_results = pd.DataFrame(df_results)
    df_output = df_params.join(df_results)
    print(df_output)
    df_output.to_csv('output-awq.tsv', sep='\t', index=False)