| from ...smp import * |
| from .multiple_choice import extract_answer_from_item |
| from PIL import Image, ImageOps |
| import numpy as np |
|
|
| FAIL_MSG = 'Failed to obtain answer via API.' |
|
|
| VQA_JUDGE_SYS_PROMPT = """ |
| You are a helpful assistant that grades answers related to visual video quality. |
| There are a lot of special terms or keywords related to video processing and photography. |
| You will pay attention to the context of `quality evaluation' when grading. |
| """ |
|
|
| VQA_JUDGE_USER_PROMPT = """ |
| Given the question {}, evaluate whether the response {} completely matches the correct answer {}. |
| First, check the response and please rate score 0 if the response is not a valid answer. |
| Please rate score 2 if the response completely or almost completely matches the correct answer on completeness, accuracy, and relevance. |
| Please rate score 1 if the response partly matches the correct answer on completeness, accuracy, and relevance. |
| Please rate score 0 if the response doesn't match the correct answer on completeness, accuracy, and relevance at all. |
| Please only provide the result in the following format: Score:' |
| """ |
|
|
|
|
| def check_ans_mcq(pred, ans, correct_choice, correct_answer): |
| flag = False |
|
|
| if correct_choice == pred or correct_choice+"." in pred or correct_answer == pred: |
| flag = True |
| elif correct_choice in pred.split("\n"): |
| flag = True |
|
|
| return flag |
|
|
| def check_ans_vqa(model, line): |
| score = model.generate(VQA_JUDGE_USER_PROMPT.format(line['question'], line['prediction'], line['answer'])).strip() |
| return score |
|
|
| def get_dimension_rating(score_file): |
| score = load(score_file) |
| result_dict = {} |
| for idx, item in score.iterrows(): |
| question_type = eval(item['dimensions'])[0].split(',')[0] |
| if question_type not in result_dict: |
| result_dict[question_type] = [0, 0] |
| result_dict[question_type][0] += int(item['score']) |
| result_dict[question_type][1] += 1 |
| return result_dict |
