無論采用直接呼叫模型測驗，還是生成模型后再去測驗，fc_dnn總是空值？求大佬救急！
報錯部分：
def classification_accuracy(Q, data_loader):
Q.eval()
labels = []
scores = []
主函式：
if __name__ == '__main__':
train_loader, valid_loader, test_loader = load_data()
best_model = 'best_DNN_model.pkl'
if not os.path.exists(os.path.join(folder, best_model)):
fc_dnn = generate_model(train_loader, valid_loader)
save_model(fc_dnn, os.path.join(folder, best_model))
else:
fc_dnn = load_model(os.path.join(folder, best_model))
test_loss, test_acc, test_auc = classification_accuracy(fc_dnn, test_loader)
print()
print('Test loss {:.4}, acc {:.4}, auc {:.4}'.format(float(test_loss), float(test_acc), float(test_auc)))
probas1, y = get_result_from_model(fc_dnn, test_loader)
average_precision = average_precision_score(y, probas1)
print('Test aupr {:.4}'.format(average_precision))
fpr, tpr, thresholds = roc_curve(y, probas1, pos_label=1)
roc_auc = auc(fpr, tpr)
print('Test auc {:.4}'.format(roc_auc))
write_result(probas1, y)
print('Done')
uj5u.com熱心網友回復：
fc_dnn總是空值？那你要除錯看看 fc_dnn = generate_model(train_loader, valid_loader)或這里 fc_dnn = load_model(os.path.join(folder, best_model))
轉載請註明出處，本文鏈接：https://www.uj5u.com/qita/221039.html
上一篇：shell 按照指定內容分割檔案