File size: 6,173 Bytes
920b3b4 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 | import os
import json
from transformers import AutoTokenizer, AutoModel
import torch
import torch.nn.functional as F
from transformers import AutoTokenizer, AutoModel
import torch
import torch.nn.functional as F
def encode_number_to_char(number):
mapping = {0: 'A', 1: 'B', 2: 'C', 3: 'D'}
return mapping.get(number, None)
#Mean Pooling - Take attention mask into account for correct averaging
def mean_pooling(model_output, attention_mask):
token_embeddings = model_output[0] #First element of model_output contains all token embeddings
input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)
txt_dir = './dev_output'
json_dir = './dev_data'
show_detail_path = './show_detail'
txt_files = {os.path.splitext(file)[0]: os.path.join(txt_dir, file) for file in os.listdir(txt_dir) if file.endswith('.txt')}
json_files = {os.path.splitext(file)[0]: os.path.join(json_dir, file) for file in os.listdir(json_dir) if file.endswith('.json')}
tokenizer = AutoTokenizer.from_pretrained('/disk6/hyhong/QWen/Sbert_pretrain') # init
model = AutoModel.from_pretrained('/disk6/hyhong/QWen/Sbert_pretrain')
question_num = 0
tau_num = 0
l3d_num = 0
neu_num = 0
huck_num = 0
india_num = 0
correct_num = 0
correct_tau_num = 0
correct_l3d_num = 0
correct_neu_num = 0
correct_huck_num = 0
correct_india_num = 0
score = []
for txt_id, txt_path in txt_files.items():
if txt_id in json_files:
json_path = json_files[txt_id]
with open(json_path, 'r', encoding='utf-8') as json_file:
json_data = json.load(json_file)
question_num += 1
if "-a-" in txt_path:
l3d_num += 1
choices = [choice[3:] for choice in json_data.get("choice", [])]
if "-b-" in txt_path:
tau_num += 1
choices = [choice[3:] for choice in json_data.get("choice", [])]
if "-c-" in txt_path:
neu_num += 1
choices = [choice[3:] for choice in json_data.get("choice", [])]
if "-d-" in txt_path:
huck_num += 1
choices = [choice[3:] for choice in json_data.get("choice", [])]
if "-e-" in txt_path:
india_num += 1
choices = [choice[3:] for choice in json_data.get("choice", [])]
with open(txt_path, 'r', encoding='utf-8') as txt_file:
txt_sentence = txt_file.read().strip()
# choices = [choice[3:] for choice in json_data.get("choice", [])]
#choices = [choice for choice in json_data.get("choice", [])]
question = json_data["question"]
answer_all = json_data["answer"]
answer = answer_all[0]
txt_name = os.path.basename(txt_path)
show_detail_txt = os.path.join(show_detail_path, txt_name)
detail_txt = open(show_detail_txt, 'w')
detail_txt.write(json_data["audio_url"])
detail_txt.write('\n')
detail_txt.write("Question:" + json_data["question"])
detail_txt.write('\n')
detail_txt.write("Choice:")
detail_txt.write('\n')
detail_txt.write(json_data["choice"][0])
detail_txt.write('\n')
if len(json_data["choice"]) == 2:
detail_txt.write(json_data["choice"][1])
detail_txt.write('\n')
if len(json_data["choice"]) == 3:
detail_txt.write(json_data["choice"][2])
detail_txt.write('\n')
if len(json_data["choice"]) == 4:
detail_txt.write(json_data["choice"][3])
detail_txt.write('\n')
detail_txt.write("Correct answer:" + answer)
detail_txt.write('\n')
new_lists = [[txt_sentence, choice] for choice in choices]
score = []
for i, new_list in enumerate(new_lists, start=1):
qwen_response = new_list[0]
new_list[0] = question + " Answer:" + new_list[0]
new_list[1] = question + " Answer:" + new_list[1]
encoded_input = tokenizer(new_list, padding=True, truncation=True, return_tensors='pt')
with torch.no_grad():
model_output = model(**encoded_input)
sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])
sentence_embeddings = F.normalize(sentence_embeddings, p=2, dim=1)
sentence_embeddings_1 = sentence_embeddings[0,:]
sentence_embeddings_2 = sentence_embeddings[1,:]
dot_product = torch.dot(sentence_embeddings_1, sentence_embeddings_2)
score.append(dot_product)
#print(f"File: {txt_id}")
#print(f"List {i}: {new_list}")
max_score = max(score)
max_index = score.index(max_score)
answer_qwen = encode_number_to_char(max_index)
detail_txt.write("Model respond:" + qwen_response)
detail_txt.write('\n')
detail_txt.write("Model answer:" + answer_qwen)
detail_txt.close()
if answer_qwen == answer:
correct_num += 1
if "-a-" in txt_path:
correct_l3d_num += 1
if "-b-" in txt_path:
correct_tau_num += 1
if "-c-" in txt_path:
correct_neu_num += 1
if "-d-" in txt_path:
correct_huck_num += 1
if "-e-" in txt_path:
correct_india_num += 1
correct_rate_all = correct_num / question_num
correct_rate_part2 = (correct_l3d_num + correct_tau_num + correct_neu_num) / (l3d_num + tau_num + neu_num)
#correct_rate_tau = correct_tau_num / tau_num
#correct_rate_neu = correct_neu_num / neu_num
correct_rate_part1 = correct_huck_num / huck_num
correct_rate_part3 = correct_india_num / india_num
print('Overall accuracy rate:')
print(correct_rate_all)
print('PART 1 accuracy rate:')
print(correct_rate_part1)
print('PART 2 accuracy rate:')
print(correct_rate_part2)
print('PART 3 accuracy rate:')
print(correct_rate_part3)
|