Keras CONV1D：檢查目標時出錯：預期解碼 output 具有形狀 (50, 50) 但得到形狀為 (50, 1) 的數組

Question

我遇到了這個問題： 檢查目標時出錯：預期 decoded_output 的形狀為 (50, 50) 但得到的數組的形狀為 (50, 1)解碼輸出）：

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)

TAM_VECTOR = X_train.shape[1]

input_tweet = Input(shape=(TAM_VECTOR,X_train.shape[2]))

encoded = Conv1D(64, kernel_size=1, activation='relu')(input_tweet)
encoded = Conv1D(32, kernel_size=1, activation='relu')(encoded)

decoded = Conv1D(32, kernel_size=1, activation='relu')(encoded)
decoded = Conv1D(64, kernel_size=1, activation='relu')(decoded)
decoded = Conv1D(TAM_VECTOR, kernel_size=1, activation='relu', name='decode_output')(decoded)

encoded = Flatten()(encoded)
second_output = Dense(1, activation='linear', name='second_output')(encoded)

autoencoder = Model(inputs=input_tweet, outputs=[decoded, second_output])

autoencoder.compile(optimizer="adam",
                    loss={'decode_output': 'mse', 'second_output': 'mse'},
                    loss_weights={'decode_output': 0.001, 'second_output': 0.999},
                    metrics=["mae"])

autoencoder.fit([X_train], [X_train, y_train], epochs=10, batch_size=32)

輸入 (X) 具有形狀 (50000,50) 但由於 Conv1D 接收到 3D 輸入，因此我將其整形為：

X = np.reshape(X, (X.shape[0], X.shape[1], -1))

(50000,50,1)

y（第二個輸出）是

y.shape

(50000,1)

這里是 model 總結

Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_43 (InputLayer)           (None, 50, 1)        0                                            
__________________________________________________________________________________________________
conv1d_169 (Conv1D)             (None, 50, 64)       128         input_43[0][0]                   
__________________________________________________________________________________________________
conv1d_170 (Conv1D)             (None, 50, 32)       2080        conv1d_169[0][0]                 
__________________________________________________________________________________________________
conv1d_171 (Conv1D)             (None, 50, 32)       1056        conv1d_170[0][0]                 
__________________________________________________________________________________________________
conv1d_172 (Conv1D)             (None, 50, 64)       2112        conv1d_171[0][0]                 
__________________________________________________________________________________________________
flatten_62 (Flatten)            (None, 1600)         0           conv1d_170[0][0]                 
__________________________________________________________________________________________________
decode_output (Conv1D)          (None, 50, 50)       3250        conv1d_172[0][0]                 
__________________________________________________________________________________________________
pib_output (Dense)              (None, 1)            1601        flatten_62[0][0]                 
==================================================================================================
Total params: 10,227
Trainable params: 10,227
Non-trainable params: 0

Answer 1

下一行中的 TAM_VECTOR 應替換為 1。

代替

decoded = Conv1D(TAM_VECTOR, kernel_size=1, activation='relu', name='decode_output')(decoded)

和

decoded = Conv1D(1, kernel_size=1, activation='relu', name='decode_output')(decoded)

解碼后的 output 形狀應與自動編碼器的輸入形狀 (50,1) 匹配。

from keras.layers import Conv1D, Flatten, Dense, Input
from keras.models import Model
import numpy as np

TAM_VECTOR = 50
input_tweet = Input(shape=(TAM_VECTOR,1))

encoded = Conv1D(64, kernel_size=1, activation='relu')(input_tweet)
encoded = Conv1D(32, kernel_size=1, activation='relu')(encoded)

decoded = Conv1D(32, kernel_size=1, activation='relu')(encoded)
decoded = Conv1D(64, kernel_size=1, activation='relu')(decoded)
decoded = Conv1D(1, kernel_size=1, activation='relu', name='decode_output')(decoded)

encoded = Flatten()(encoded)
second_output = Dense(1, activation='linear', name='second_output')(encoded)

autoencoder = Model(inputs=input_tweet, outputs=[decoded, second_output])

autoencoder.compile(optimizer="adam",
                    loss={'decode_output': 'mse', 'second_output': 'mse'},
                    loss_weights={'decode_output': 0.001, 'second_output': 0.999},
                    metrics=["mae"])

autoencoder.fit(np.ones((1,50,1)), [np.ones((1,50,1)), np.ones((1,1))])

1/1 [==============================] - 0s 407ms/步 - 損失：0.9112 - decode_output_loss：0.9549 - second_output_loss ：0.9111 - decode_output_mean_absolute_error：0.9772 - second_output_mean_absolute_error：0.9545

Answer 2

這是錯誤：錯誤1）：

InvalidArgumentError: 2 root error(s) found.
(0) Invalid argument: You must feed a value for placeholder tensor 'decode_output_sample_weights_32' with dtype float and shape [?]
[[{{node decode_output_sample_weights_32}}]]
 [[loss_2/second_output_loss/Mean_3/_3217]]
  (1) Invalid argument: You must feed a value for placeholder tensor 'decode_output_sample_weights_32' with dtype float and shape [?]
     [[{{node decode_output_sample_weights_32}}]]
0 successful operations.
0 derived errors ignored.

錯誤2）：

InvalidArgumentError: You must feed a value for placeholder tensor 'decode_output_target_17' with dtype float and shape [?,?,?] [[{{node decode_output_target_17}}]]

錯誤3）：

UnknownError: 2 root error(s) found.
  (0) Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
     [[{{node conv1d_1/convolution}}]]
     [[loss/add/_157]]
  (1) Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
     [[{{node conv1d_1/convolution}}]]
0 successful operations.
0 derived errors ignored.

錯誤 4):

UnknownError: 2 root error(s) found.
  (0) Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
     [[{{node conv1d_25/convolution}}]]
     [[loss_6/second_output_loss/Mean_3/_1025]]
  (1) Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
     [[{{node conv1d_25/convolution}}]]
0 successful operations.
0 derived errors ignored.