如何将标准标准器正确加载到TensorFlow Keras型号？

发布于 2025-02-11 02:29:33 字数 1110 浏览 1 评论 0原文

我对TensorFlow模型的负载标准标准器有问题。我使用以下代码来加载标准标准模型：

scaler_load = pickle.load(open(path + save_dir +'std_scaler_1.pkl', 'rb'))
X_test_load = scaler_load.transform(X_test_fs)
X_test_load

但是，当我从磁盘加载模型并使用此代码进行编译时：

load_model = tf.keras.models.load_model(path + save_dir + 'mlp_model_fs_25_1.h5')
load_model.compile(loss=[tf.keras.losses.CategoricalCrossentropy(), tf.keras.losses.MeanSquaredError()], 
              optimizer=tf.keras.optimizers.Adam(learning_rate=0.0001), 
              metrics=['accuracy', tf.keras.metrics.MeanSquaredError()])
load_model.evaluate(X_test_load, y_test)

结果与我进行第一次培训和测试时不同。当第一次的结果如下时：

54/54 [==============================] - 0s 2ms/step - loss: 0.1129 - accuracy: 0.9662 - mean_squared_error: 0.0294
[0.11291049420833588, 0.9661807417869568, 0.02943398430943489]

但是当我加载它时，分数就是这样

54/54 [==============================] - 0s 1ms/step - loss: 0.4714 - accuracy: 0.8583 - mean_squared_error: 0.1088
[0.47135162353515625, 0.8583090305328369, 0.10878879576921463]

原文

i have a problem about load StandardScaler to my tensorflow model.
I used the following code to load the StandardScaler model :

scaler_load = pickle.load(open(path + save_dir +'std_scaler_1.pkl', 'rb'))
X_test_load = scaler_load.transform(X_test_fs)
X_test_load

But when i load the model from disk and compile it with this code :

load_model = tf.keras.models.load_model(path + save_dir + 'mlp_model_fs_25_1.h5')
load_model.compile(loss=[tf.keras.losses.CategoricalCrossentropy(), tf.keras.losses.MeanSquaredError()], 
              optimizer=tf.keras.optimizers.Adam(learning_rate=0.0001), 
              metrics=['accuracy', tf.keras.metrics.MeanSquaredError()])
load_model.evaluate(X_test_load, y_test)

The results are different from when I did the first training and testing. When the first time, the results are as follows:

54/54 [==============================] - 0s 2ms/step - loss: 0.1129 - accuracy: 0.9662 - mean_squared_error: 0.0294
[0.11291049420833588, 0.9661807417869568, 0.02943398430943489]

But when i load it the score is look like this

54/54 [==============================] - 0s 1ms/step - loss: 0.4714 - accuracy: 0.8583 - mean_squared_error: 0.1088
[0.47135162353515625, 0.8583090305328369, 0.10878879576921463]

分享到QQ

分享到微博