Tensorflow 与层不兼容：预期 min_ndim=3，发现 ndim=2。错误
Tensorflow incompatible with the layer: expected min_ndim=3, found ndim=2. error
我有所需形状的数据集，但 Tensorflow CNN 模型不接受它。我不知道我应该改变什么。
train_X.shape, train_Y_one_hot.shape, test_X.shape, test_Y_one_hot.shape
((1073965, 15), (1073965, 8), (268492, 15), (268492, 8))
train_X[0]
array([-3.75609105e-06, 7.77089567e-05, 4.41936374e-05, -6.78777110e-07,
1.17593595e-04, -7.11814884e-05, 8.92277296e-05, 5.97469947e-05,
-2.17992064e-05, -1.38704555e-05, 1.66007993e-05, 6.45235327e-05,
4.69597312e-06, 5.03974802e-06, 4.71866744e-04])
CNN 模型：
def classifySupervised(train_X, train_Y_one_hot, test_X, test_Y_one_hot,valid_X, valid_Y_one_hot,num_classes):
# train_data = tf.data.Dataset.from_tensor_slices((train_X,train_Y_one_hot))
# valid_data = tf.data.Dataset.from_tensor_slices((valid_X,valid_Y_one_hot))
batch_size = 21
epochs = 30
#Complete stock code
model1 = Sequential()
model1.add(Conv1D(32, kernel_size=3,activation='relu',input_shape=(15),padding='same'))
model1.add(LeakyReLU(alpha=0.1))
model1.add(MaxPooling1D(2,padding='same'))
model1.add(Dropout(0.25))
model1.add(Conv1D(64, 3, activation='relu',padding='same'))
model1.add(LeakyReLU(alpha=0.1))
model1.add(MaxPooling1D(pool_size=2,padding='same'))
model1.add(Dropout(0.25))
model1.add(Conv1D(128, 3, activation='relu',padding='same'))
model1.add(LeakyReLU(alpha=0.20))
model1.add(MaxPooling1D(pool_size=2,padding='same'))
model1.add(Dropout(0.4))
model1.add(Dense(128, activation='relu'))
model1.add(LeakyReLU(alpha=0.1))
model1.add(Dropout(0.3))
model1.add(Dense(num_classes, activation='softmax'))
model1.compile(loss=keras.losses.categorical_crossentropy, optimizer=tf.keras.optimizers.Adam(),metrics=['accuracy'])
model1_train_dropout = model1.fit(train_X,train_Y_one_hot, batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(valid_X,valid_Y_one_hot))
model1.save("Blythe_gt_train.h5py")
#loadedmodel = .load("sanjoukin_gt_train.h5py")
test_eval = model1.evaluate(test_X, test_Y_one_hot, verbose=11)
#Prediction maybe sepearate function
predicted_classes = model1.predict(test_X)
predicted_classes = np.argmax(np.round(predicted_classes),axis=1)
predicted_classes.shape, test_Y.shape
correct = np.where(predicted_classes==test_Y)[0]
print("Found %d correct labels" % len(correct))
for i, correct in enumerate(correct[:9]):
plt.subplot(3,3,i+1)
plt.imshow(test_X[correct].reshape(28,28), cmap='gray', interpolation='none')
plt.title("Predicted {}, Class {}".format(predicted_classes[correct], test_Y[correct]))
plt.tight_layout()
incorrect = np.where(predicted_classes!=test_Y)[0]
print("Found %d incorrect labels" % len(incorrect))
for i, incorrect in enumerate(incorrect[:9]):
plt.subplot(3,3,i+1)
plt.imshow(test_X[incorrect].reshape(28,28), cmap='gray', interpolation='none')
plt.title("Predicted {}, Class {}".format(predicted_classes[incorrect], test_Y[incorrect]))
plt.tight_layout()
#Classification Report
target_names = ["Class {}".format(i) for i in range(num_classes)]
print(classification_report(test_Y, predicted_classes, target_names=target_names))
return test_eval
pass
classifySupervised(train_X, train_Y_one_hot, test_X, test_Y_one_hot,valid_X,valid_Y_one_hot,nClasses)
这是我得到的错误：
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
Input In [41], in <cell line: 56>()
54 return test_eval
55 pass
---> 56 classifySupervised(train_X, train_Y_one_hot, test_X, test_Y_one_hot,valid_X,valid_Y_one_hot,nClasses)
Input In [41], in classifySupervised(train_X, train_Y_one_hot, test_X, test_Y_one_hot, valid_X, valid_Y_one_hot, num_classes)
6 #Complete stock code
7 model1 = Sequential()
----> 8 model1.add(Conv1D(32, kernel_size=3,activation='relu',input_shape=(15,),padding='same'))
9 model1.add(LeakyReLU(alpha=0.1))
10 model1.add(MaxPooling1D(2,padding='same'))
File ~\.conda\envs\tf3\lib\site-packages\tensorflow\python\training\tracking\base.py:587, in no_automatic_dependency_tracking.<locals>._method_wrapper(self, *args, **kwargs)
585 self._self_setattr_tracking = False # pylint: disable=protected-access
586 try:
--> 587 result = method(self, *args, **kwargs)
588 finally:
589 self._self_setattr_tracking = previous_value # pylint: disable=protected-access
File ~\.conda\envs\tf3\lib\site-packages\keras\utils\traceback_utils.py:67, in filter_traceback.<locals>.error_handler(*args, **kwargs)
65 except Exception as e: # pylint: disable=broad-except
66 filtered_tb = _process_traceback_frames(e.__traceback__)
---> 67 raise e.with_traceback(filtered_tb) from None
68 finally:
69 del filtered_tb
File ~\.conda\envs\tf3\lib\site-packages\keras\engine\input_spec.py:228, in assert_input_compatibility(input_spec, inputs, layer_name)
226 ndim = x.shape.rank
227 if ndim is not None and ndim < spec.min_ndim:
--> 228 raise ValueError(f'Input {input_index} of layer "{layer_name}" '
229 'is incompatible with the layer: '
230 f'expected min_ndim={spec.min_ndim}, '
231 f'found ndim={ndim}. '
232 f'Full shape received: {tuple(shape)}')
233 # Check dtype.
234 if spec.dtype is not None:
ValueError: Input 0 of layer "conv1d_8" is incompatible with the layer: expected min_ndim=3, found ndim=2. Full shape received: (None, 15)
train_X 和 test_X 缺少维度。 Conv1D 层需要具有形状（样本、时间步长、特征）的数据。因此，您可以做的最简单的事情是在调用 model.fit(*):
之前添加一个维度
train_X = tf.expand_dims(train_X, axis=-1)
test_X = tf.expand_dims(test_X, axis=-1)
此外，将模型中的最后一个 MaxPooling1D 层替换为 GlobalMaxPool1D 层。
MaxPooling1D 层针对给定的 window 大小和步幅计算张量的最大值。然后它输出一个 3D 张量 (batch_size, timesteps, features)，而 Dense 层通常使用 2D 张量 (batch_size, features)。 GlobalMaxPool1D 层通过取时间维度上的最大值和 returns 二维张量对输入表示进行下采样。
我有所需形状的数据集，但 Tensorflow CNN 模型不接受它。我不知道我应该改变什么。
train_X.shape, train_Y_one_hot.shape, test_X.shape, test_Y_one_hot.shape
((1073965, 15), (1073965, 8), (268492, 15), (268492, 8))
train_X[0]
array([-3.75609105e-06, 7.77089567e-05, 4.41936374e-05, -6.78777110e-07,
1.17593595e-04, -7.11814884e-05, 8.92277296e-05, 5.97469947e-05,
-2.17992064e-05, -1.38704555e-05, 1.66007993e-05, 6.45235327e-05,
4.69597312e-06, 5.03974802e-06, 4.71866744e-04])
CNN 模型：
def classifySupervised(train_X, train_Y_one_hot, test_X, test_Y_one_hot,valid_X, valid_Y_one_hot,num_classes):
# train_data = tf.data.Dataset.from_tensor_slices((train_X,train_Y_one_hot))
# valid_data = tf.data.Dataset.from_tensor_slices((valid_X,valid_Y_one_hot))
batch_size = 21
epochs = 30
#Complete stock code
model1 = Sequential()
model1.add(Conv1D(32, kernel_size=3,activation='relu',input_shape=(15),padding='same'))
model1.add(LeakyReLU(alpha=0.1))
model1.add(MaxPooling1D(2,padding='same'))
model1.add(Dropout(0.25))
model1.add(Conv1D(64, 3, activation='relu',padding='same'))
model1.add(LeakyReLU(alpha=0.1))
model1.add(MaxPooling1D(pool_size=2,padding='same'))
model1.add(Dropout(0.25))
model1.add(Conv1D(128, 3, activation='relu',padding='same'))
model1.add(LeakyReLU(alpha=0.20))
model1.add(MaxPooling1D(pool_size=2,padding='same'))
model1.add(Dropout(0.4))
model1.add(Dense(128, activation='relu'))
model1.add(LeakyReLU(alpha=0.1))
model1.add(Dropout(0.3))
model1.add(Dense(num_classes, activation='softmax'))
model1.compile(loss=keras.losses.categorical_crossentropy, optimizer=tf.keras.optimizers.Adam(),metrics=['accuracy'])
model1_train_dropout = model1.fit(train_X,train_Y_one_hot, batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(valid_X,valid_Y_one_hot))
model1.save("Blythe_gt_train.h5py")
#loadedmodel = .load("sanjoukin_gt_train.h5py")
test_eval = model1.evaluate(test_X, test_Y_one_hot, verbose=11)
#Prediction maybe sepearate function
predicted_classes = model1.predict(test_X)
predicted_classes = np.argmax(np.round(predicted_classes),axis=1)
predicted_classes.shape, test_Y.shape
correct = np.where(predicted_classes==test_Y)[0]
print("Found %d correct labels" % len(correct))
for i, correct in enumerate(correct[:9]):
plt.subplot(3,3,i+1)
plt.imshow(test_X[correct].reshape(28,28), cmap='gray', interpolation='none')
plt.title("Predicted {}, Class {}".format(predicted_classes[correct], test_Y[correct]))
plt.tight_layout()
incorrect = np.where(predicted_classes!=test_Y)[0]
print("Found %d incorrect labels" % len(incorrect))
for i, incorrect in enumerate(incorrect[:9]):
plt.subplot(3,3,i+1)
plt.imshow(test_X[incorrect].reshape(28,28), cmap='gray', interpolation='none')
plt.title("Predicted {}, Class {}".format(predicted_classes[incorrect], test_Y[incorrect]))
plt.tight_layout()
#Classification Report
target_names = ["Class {}".format(i) for i in range(num_classes)]
print(classification_report(test_Y, predicted_classes, target_names=target_names))
return test_eval
pass
classifySupervised(train_X, train_Y_one_hot, test_X, test_Y_one_hot,valid_X,valid_Y_one_hot,nClasses)
这是我得到的错误：
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
Input In [41], in <cell line: 56>()
54 return test_eval
55 pass
---> 56 classifySupervised(train_X, train_Y_one_hot, test_X, test_Y_one_hot,valid_X,valid_Y_one_hot,nClasses)
Input In [41], in classifySupervised(train_X, train_Y_one_hot, test_X, test_Y_one_hot, valid_X, valid_Y_one_hot, num_classes)
6 #Complete stock code
7 model1 = Sequential()
----> 8 model1.add(Conv1D(32, kernel_size=3,activation='relu',input_shape=(15,),padding='same'))
9 model1.add(LeakyReLU(alpha=0.1))
10 model1.add(MaxPooling1D(2,padding='same'))
File ~\.conda\envs\tf3\lib\site-packages\tensorflow\python\training\tracking\base.py:587, in no_automatic_dependency_tracking.<locals>._method_wrapper(self, *args, **kwargs)
585 self._self_setattr_tracking = False # pylint: disable=protected-access
586 try:
--> 587 result = method(self, *args, **kwargs)
588 finally:
589 self._self_setattr_tracking = previous_value # pylint: disable=protected-access
File ~\.conda\envs\tf3\lib\site-packages\keras\utils\traceback_utils.py:67, in filter_traceback.<locals>.error_handler(*args, **kwargs)
65 except Exception as e: # pylint: disable=broad-except
66 filtered_tb = _process_traceback_frames(e.__traceback__)
---> 67 raise e.with_traceback(filtered_tb) from None
68 finally:
69 del filtered_tb
File ~\.conda\envs\tf3\lib\site-packages\keras\engine\input_spec.py:228, in assert_input_compatibility(input_spec, inputs, layer_name)
226 ndim = x.shape.rank
227 if ndim is not None and ndim < spec.min_ndim:
--> 228 raise ValueError(f'Input {input_index} of layer "{layer_name}" '
229 'is incompatible with the layer: '
230 f'expected min_ndim={spec.min_ndim}, '
231 f'found ndim={ndim}. '
232 f'Full shape received: {tuple(shape)}')
233 # Check dtype.
234 if spec.dtype is not None:
ValueError: Input 0 of layer "conv1d_8" is incompatible with the layer: expected min_ndim=3, found ndim=2. Full shape received: (None, 15)
train_X 和 test_X 缺少维度。 Conv1D 层需要具有形状（样本、时间步长、特征）的数据。因此，您可以做的最简单的事情是在调用 model.fit(*):
train_X = tf.expand_dims(train_X, axis=-1)
test_X = tf.expand_dims(test_X, axis=-1)
此外，将模型中的最后一个 MaxPooling1D 层替换为 GlobalMaxPool1D 层。
MaxPooling1D 层针对给定的 window 大小和步幅计算张量的最大值。然后它输出一个 3D 张量 (batch_size, timesteps, features)，而 Dense 层通常使用 2D 张量 (batch_size, features)。 GlobalMaxPool1D 层通过取时间维度上的最大值和 returns 二维张量对输入表示进行下采样。