import json import string import regex #Normalization from SQuAD evaluation script https://worksheets.codalab.org/rest/bundles/0x6b567e1cf2e041ec80d7098f031c5c9e/contents/blob/ def normalize_answer(s): def remove_articles(text): return regex.sub(r'\b(a|an|the)\b', ' ', text) def white_space_fix(text): return ' '.join(text.split()) def remove_punc(text): exclude = set(string.punctuation) return ''.join(ch for ch in text if ch not in exclude) def lower(text): return text.lower() return white_space_fix(remove_articles(remove_punc(lower(s)))) def cal_acc_multi(ground_truth, preds, return_id = False): all_num = len(ground_truth) acc_num = 0 ids = [] temp = [] for i, answer_id in enumerate(ground_truth): pred = preds[i] cnt = 0 for aid in answer_id: if pred == aid: cnt += 1 if cnt ==1: acc_num += 1/3 elif cnt == 2: acc_num += 2/3 elif cnt > 2: acc_num += 1 if return_id: return acc_num / all_num, ids else: return acc_num, all_num def ensemble(a): return max(a[::-1], key = a.count) # Ground Truth Answers f=open("/root/okvqa/data/okvqa_val.json", "r") answer_dict=json.load(f) f.close() for k in answer_dict.keys(): for a_ind, a in enumerate(answer_dict[k]['multi_answers']): answer_dict[k]['multi_answers'][a_ind] = normalize_answer(answer_dict[k]['multi_answers'][a_ind]) # Load Predictions (for example, ensemble of three models' predictions) f1=open("/mnt/bn/qingyi-hl/finetunedModelonOKVQA/1e-41e-5FTwiki25-From-1e-41e-5PretrainWiki25Epo0/FTwiki25FromPretrainWiki25Epo0-1e41e5/predictions.json", "r") predict0_dict=json.load(f1) for p in predict0_dict.keys(): predict0_dict[p]=normalize_answer(predict0_dict[p]) f1.close() f2=open("/mnt/bn/qingyi-hl/finetunedModelonOKVQA/1e-41e-5FTwiki25-From-1e-41e-5PretrainWiki25Epo1/predictions.json", "r") predict1_dict=json.load(f2) for p in predict1_dict.keys(): predict1_dict[p]=normalize_answer(predict1_dict[p]) f2.close() f3=open("/mnt/bn/qingyi-hl/finetunedModelonOKVQA/1e-41e-5FTwiki25-From-1e-41e-5PretrainWiki25Epo2/predictions.json", "r") predict2_dict=json.load(f3) for p in predict2_dict.keys(): predict2_dict[p]=normalize_answer(predict2_dict[p]) f3.close() answer_list=[] predict0_list=[] predict1_list=[] predict2_list=[] emsemble_predict=[] for k in answer_dict.keys(): answer_list.append( answer_dict[k]['multi_answers']) predict0_list.append( predict0_dict[k]) predict1_list.append( predict1_dict[k]) predict2_list.append( predict2_dict[k]) emsemble_predict.append(ensemble([predict0_dict[k], predict1_dict[k], predict2_dict[k]) acc_n0,all_n0=cal_acc_multi(answer_list,predict0_list) acc_n1,all_n1=cal_acc_multi(answer_list,predict1_list) acc_n2,all_n2=cal_acc_multi(answer_list,predict2_list) acc_ens,all_ens=cal_acc_multi(answer_list,emsemble_predict) print("0-accuracy",acc_n0/all_n0) print("1-accuracy",acc_n1/all_n1) print("2-accuracy",acc_n2/all_n2) print("ensemble-accuracy",acc_ens/all_ens)