data_array.pkl is sentence transformers for the questions.pkl
import pickle
with open('data_array.pkl', 'rb') as f: # Open file in read-binary mode
loaded_data = pickle.load(f)
Accessing data in the benchmark.pkl
all_comparisons = bds[168]['compare']
just_one_response = bds[168]['compare']['llamallama2-70b']
original_prompt = bds[168]['conversation'][0]['content']
original_response = bds[168]['conversation'][1]['content']