1 predictions = self.model.predict(x_data, verbose=0)[0] 2 y_pred_idx = np.argmax(predictions) 3 y_pred_prob = predictions[y_pred_idx] 4 y_pred_label = self.le.inverse_transform(y_pred_idx) 5 res = OrderedDict() 6 res['first'] = self.first_label2name[y_pred_label[0]] 7 res['second'] = self.second_label2name[y_pred_label] 8 res['probability'] = y_pred_prob 9 out.write(json.dumps(res, ensure_ascii=False) + '\n')
報錯：'0.80454153 is not JSON serializable'
輸出y_pred_prob的類別：<type 'numpy.float32'>
參考https://stackoverflow.com/questions/27050108/convert-numpy-type-to-python/27050186#27050186中類似問題的解決方案，有個回答中提到：
Ultimately, it looks like json is telling you that an int isn't serializable, but really, it's telling you that this particular np.int32 (or whatever type you actually have) isn't serializable. (No real surprise there -- No np.int32 is serializable).
可以知道，json不能對np.int32或者np.float32等類型進行序列化，可以通過自定義serializer或者直接類型轉換來解決。
因此，在第８行顯式地把數據轉為float類型：res['probability'] = float(y_pred_prob)即可