| import torch |
| import os |
| import json |
| import gc |
| import csv |
|
|
| from PIL import Image |
| from datasets import load_dataset,load_from_disk |
| from collections import defaultdict |
| from tqdm import tqdm |
|
|
| from ..utils import save_json,extract,judge_multi_choice |
| from ..base_dataset import BaseDataset |
|
|
| from ..question_formats import get_multiple_choice_prompt |
| medical_subject = ["anatomy","clinical_knowledge","college_biology","college_medicine","medical_genetics","professional_medicine","nutrition","virology","high_school_biology"] |
|
|
| class MMLU(BaseDataset): |
| def __init__(self,model,dataset_path,output_path,subject = "all"): |
| self.model = model |
| self.output_path = output_path |
| self.dataset_path = dataset_path if dataset_path else "cais/mmlu" |
| self.samples = [] |
| self.subject = subject |
| self.chunk_idx = int(os.environ.get("chunk_idx",0)) |
| self.num_chunks = int(os.environ.get("num_chunks",1)) |
| |
| def load_data(self): |
| dataset = [] |
| if self.subject == "all": |
| datas = load_dataset(self.dataset_path,"all")["test"] |
| for data in datas: |
| dataset.append(data) |
| elif self.subject == "medical": |
| for split in medical_subject: |
| datas = load_dataset(self.dataset_path,split)["test"] |
| for data in datas: |
| dataset.append(data) |
|
|
| for idx,sample in tqdm(enumerate(dataset)): |
| if idx % self.num_chunks == self.chunk_idx: |
| sample = self.construct_messages(sample) |
| self.samples.append(sample) |
| return self.samples |
|
|
| def construct_messages(self,sample): |
| choices = sample["choices"] |
| answer = sample["answer"] |
| question = sample["question"] |
| alphas = ["A","B","C","D"] |
|
|
| choices = "\n".join([f"{alpha}.{choice}" for alpha,choice in zip(alphas,choices)]) |
| is_reasoning = True if os.environ.get("REASONING","False") == "True" else False |
| prompt = f"Context: {context}" + "\n" + get_multiple_choice_prompt(question,choices,is_reasoning) |
|
|
| messages = {"prompt":prompt} |
| sample["prompt"] = prompt |
| sample["messages"] = messages |
| sample["answer"] = alphas[answer] |
| return sample |
|
|
|
|
| def cal_metrics(self,out_samples): |
| total = 0 |
| right = 0 |
| for i,sample in enumerate(out_samples): |
| response = sample["response"] |
| response = extract(response,"answer") |
| choices = sample["choices"] |
| answer = sample["answer"] |
|
|
| correct = judge_multi_choice(choices,answer,response) |
| out_samples[i]["correct"] = correct |
| if correct: |
| right += 1 |
| total += 1 |
|
|
| metrics = {"total metrics":{"total":total,"right":right,"acc":right/total}} |
| return metrics,out_samples |
|
|
|
|
|
|
| |
