RLHFlow/Orca-distibalel-standard

dataset_info:
features:

• name: chosen
  list:
  • name: content
    dtype: string
  • name: role
    dtype: string
• name: rejected
  list:
  • name: content
    dtype: string
  • name: role
    dtype: string
• name: chosen_score
  dtype: float64
• name: rejected_score
  dtype: float64
  splits:
• name: train
  num_bytes: 23664435.680068433
  num_examples: 6926
  download_size: 12516825
  dataset_size: 23664435.680068433
  configs:
• config_name: default
  data_files:
  • split: train
    path: data/train-*

From argilla/distilabel-intel-orca-dpo-pairs. We delete the pairs with the same score.

def process_example(example):
    prompt = example['input']
    chosen_response = example['chosen']
    rejected_response = example['rejected']
    if example['rating'] is None:
        chosen_rating = None
        rejected_rating = None
    else:
        if example['order'][0] == 'chosen':
            chosen_rating = example['rating'][0]
            rejected_rating = example['rating'][1]
        elif example['order'][0] == 'rejected':
            chosen_rating = example['rating'][1]
            rejected_rating = example['rating'][0]
        else:
            print('error')
        
    chosen_message = [
        {"content": prompt, "role": "user"},
                {"content": chosen_response, "role": "assistant"},
    ]
    rejected_message = [
        {"content": prompt, "role": "user"},
                {"content": rejected_response, "role": "assistant"},
    ]
    
    
    return {"rejected": rejected_message, "chosen": chosen_message, "rejected_score": rejected_rating, "chosen_score": chosen_rating}