LLM_Evaluator/evaluators/chatglm_mixin.py

import os
import re
from tqdm import tqdm
import torch
from transformers import AutoTokenizer, AutoModel, AutoConfig
from transformers.generation.logits_process import LogitsProcessor
from transformers.generation.utils import LogitsProcessorList
from evaluators.evaluator import Evaluator

class ChatGLMMixin:
    def __init__(self):
        self.tokenizer = None
        self.model = None
        self.model_name = None
        self.k = None
        self.choices = None
        self.finetune_name = None

    def eval_subject(self, subject_name, test_df, dev_df=None, few_shot=False, cot=False, save_result_dir=None):
        correct_num = 0
        result = []
        score = []
        answer_list = []
        if few_shot:
            history = self.generate_few_shot_prompt(subject_name, dev_df, cot=cot)
        else:
            history = self.generate_zero_shot_prompt(is_choice_question=True)
        answers = list(test_df['answer'])
        for row_index, row in tqdm(test_df.iterrows(), total=len(test_df)):
            question = self.format_example(row, include_answer=False, cot=cot)
            if few_shot:
                response, _ = self.model.chat(self.tokenizer, question, max_length=2000,
                                              do_sample=False, history=history)
                response = response.strip()
                ans, direct_extract = self.extract_cot_answer(row, response)
            else:  # zero-shot by extracting answer from distribution
                response, _ = self.model.chat(self.tokenizer, question, max_length=2000,
                                              do_sample=False, history=history)
                response = response.strip()
                ans, direct_extract = self.extract_cot_answer(row, response)
            if ans == answers[row_index]:
                correct_num += 1
                correct = 1
            else:
                correct = 0
            if save_result_dir:
                result.append(response)
                score.append(correct)
                answer_list.append(ans)
        correct_ratio = 100 * correct_num / len(answers)

        if save_result_dir:
            test_df['model_output'] = result
            test_df['correctness'] = score
            test_df['model_answer'] = answer_list
            result_file_name = f'{subject_name}_{correct_ratio}_test.csv'
            if few_shot:
                result_file_name = f'{subject_name}_{correct_ratio}_few_shot_test.csv'
            test_df.to_csv(os.path.join(save_result_dir, result_file_name))

        return correct_ratio

    def eval_qa(self, subject_name, qa_df, save_result_dir=None):
        history = self.generate_zero_shot_prompt(is_choice_question=False)
        for row_index, row in tqdm(qa_df.iterrows(), total=len(qa_df)):
            question = row['question']
            response, _ = self.model.chat(self.tokenizer, question, max_length=2000,
                                          do_sample=False, history=history)
            response = response.strip()
            qa_df.loc[row_index, 'model_output'] = response
            # current_length = 0
            # response = ""
            # for resp, _ in self.model.stream_chat(self.tokenizer, question, max_length=300,
            #                                       do_sample=False, history=history):
            #     print(resp[current_length:], end="", flush=True)
            #     current_length = len(resp)
            #     response = resp
            # print('')
        if save_result_dir and self.finetune_name is not None:
            result_file_name = f'{subject_name}_qa_test_result.csv'
            qa_df.to_csv(os.path.join(save_result_dir, result_file_name))
        return qa_df

    def generate_few_shot_prompt(self, subject, dev_df, cot=False):
        message = []
        k = self.k
        if self.k == -1:
            k = dev_df.shape[0]
        init_example = self.format_example(dev_df.iloc[0, :], cot=cot,
                                           add_prompt=f"以下是中国关于{subject}考试的单项选择题，请选出其中的正确答案。\n\n")
        if isinstance(init_example, list):
            message.extend(init_example)
        else:
            message.append(init_example)
        for i in range(1, k):
            example = self.format_example(dev_df.iloc[i, :], cot=cot)
            if isinstance(example, list):
                message.extend(example)
            else:
                message.append(example)
        return message

    def generate_zero_shot_prompt(self, is_choice_question=True):
        if self.model_name == 'chatglm3' and is_choice_question:
            return [{'role': 'user',
                     'content': '接下来会提供给你一些选择题，请选出正确的答案，给出正确的选项即可。'},
                    {'role': 'assistant',
                     'content': '好的，我会尽力解答。'}]
        elif self.model_name == 'chatglm3' and not is_choice_question:
            return [{'role': 'user',
                     'content': '接下来会给你一些一些汽车领域相关问题，请回答。'},
                    {'role': 'assistant',
                     'content': '好的，我会尽力解答。'}]
        else:
            return []

    def format_example(self, line, include_answer=True, cot=False, add_prompt=''):
        example = add_prompt + line['question']
        # print(example)
        for choice in self.choices:
            example += f'\n{choice}. {line[f"{choice}"]}'
        example += '\n答案：'
        if include_answer:
            if cot:
                ans = "让我们一步一步思考，\n" + line["explanation"] + f"\n所以答案是{line['answer']}。"
            else:
                ans = line["answer"]
            if self.model_name == 'chatglm3':
                m = [{
                    'role': 'user',
                    'content': example
                }, {
                    'role': 'assistant',
                    'content': ans
                }]
            else:
                m = (example, ans)
            return m
        return example

    def extract_cot_answer(self, line, gen_ans):
        m = re.findall(r'所以答案是(.+?)。', gen_ans, re.M)
        if len(m) > 0 and m[-1] in self.choices:
            return m[-1], True
        answer_patterns = [
            r'([ABCD])是正确的',
            r'选项([ABCD])正确',
            r'答案为([ABCD])',
            r'答案是([ABCD])',
            r'答案([ABCD])',
            r'选择([ABCD])',
            r'答案：([ABCD])',
            r'选择答案([ABCD])',
            r'正确答案是([ABCD])'
        ]
        # RE extraction
        for answer_pattern in answer_patterns:
            m = re.search(answer_pattern, gen_ans, re.M)
            if m:
                answer = m.group(1)
                return answer, False
        # only containing one choice-character
        m = re.findall(r'[ABCD]', gen_ans, re.M)
        if len(m) == 1:
            answer = m[0]
            return answer, False
        answer_word_counter = 0
        # only containing one choice-context
        for c in self.choices:
            if str(line[f'{c}']) in gen_ans:
                answer = c
                answer_word_counter += 1
        if answer_word_counter == 1:
            return answer, False
        return '-', False