我的预训练模型的输入形状是 (None, 1, 32),应用迁移学习的目标模型有一个输入形状 (None, 1, 128)。如何更改输入形状? 我的代码:
loaded_model = load_model('/ANN NS/ANN_NS_1ch.h5')
# load the dataset
dataset = pd.read_csv(r'C:\Users\ABC\.spyder-py3\abc.csv', header = None, )
print(dataset.shape)
your text`dataset = dataset.iloc[1:,:]
df = dataset.iloc[:1100,38:167]
Y= dataset.iloc[:1100,1:37]
#train_test_split
X_train1,X_test1,Y_train1,Y_test1 = train_test_split(X,Y,test_size=0.2,random_state=None,shuffle=False)
Y_train_log = np.log(Y_train1)
Y_test_log = np.log(Y_test1)
X_tr= X_train1.values.reshape(X_train1.shape[0],1,X_train1.shape[1])
X_ts =X_test1.values.reshape(X_test1.shape[0],1,X_test1.shape[1])
t1=time.time()
ANN = tf.keras.models.Sequential()
new_model = loaded_model.layers
for layer in loaded_model.layers[:-1]: ANN.add(layer)
for layer in loaded_model.layers: layer.trainable = False
tf.random.set_seed(71)
ANN.add(tf.keras.layers.Dense(100, input_shape= (1,128), activation ='tanh')
错误:
ValueError:“顺序”层的输入 0 与 图层:预期形状=(None, 1, 32),发现形状=(None, 1, 128)
我想通过使用 32 输入形状预训练模型将形状从 32 更改为 128。