ValueError: 层序贯_1 的输入 0 与层不兼容：预期 ndim=3，发现 ndim=2。收到的完整形状：（无，93）

Posted 2023-02-16

【中文标题】ValueError: 层序贯_1 的输入 0 与层不兼容：预期 ndim=3，发现 ndim=2。收到的完整形状：（无，93）

【英文标题】：ValueError: Input 0 of layer sequential_1 is incompatible with the layer: expected ndim=3, found ndim=2. Full shape received: (None, 93)

【发布时间】：2021-08-25 23:57:49

【问题描述】：

我想通过时间序列让 LSTM 进行预测，但我犯了这个错误。

我的 X_train 和 y_train 形状

X_train_seasonal.shape

(893, 93)

y_train_seasonal.shape

(893,)

我的 LSTM

def getModel():
    model = tf.keras.Sequential()
    model.add(tf.keras.layers.LSTM(100, activation='relu', input_shape=(X_train_seasonal.shape[0],X_train_seasonal[1])))
    model.add(tf.keras.layers.Dense(units=1))
    return model

我的模型编译：

model = getModel()
model.compile(
  loss='mean_squared_error',
  optimizer=tf.keras.optimizers.Adam(0.001)
)

我的历史：

history_seasonal = model.fit(
    X_train_seasonal, y_train_seasonal,
    epochs=100,
    batch_size=32,
    validation_split=0.1,
    shuffle=False
)

我得到了那个错误：

ValueError: 层序号_1 的输入 0 与层不兼容：预期 ndim=3，发现 ndim=2。收到的完整形状：（无，93）

我是 LSTM 的新手，如果有任何帮助，我将不胜感激。谢谢大家。

【问题讨论】：

LSTM 层需要一个时间维度，因此您输入的格式应为[batch, timesteps, feature]。

谢谢帕维尔，这真的很有帮助。我在我的 x 训练和测试数据集中进行了重塑并开始工作。非常感谢

【参考方案1】：

我能够使用示例代码复制您的问题，如下所示

import tensorflow as tf
import numpy as np

inputs = tf.random.normal([10, 8])
simple_lstm = tf.keras.layers.LSTM(4)
output = simple_lstm(inputs) 
print(output)

输出：

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-1-2cb62bec2981> in <module>()
      4 inputs = tf.random.normal([10, 8])
      5 simple_lstm = tf.keras.layers.LSTM(4)
----> 6 output = simple_lstm(inputs)
      7 print(output)

2 frames
/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/input_spec.py in assert_input_compatibility(input_spec, inputs, layer_name)
    217                          'expected ndim=' + str(spec.ndim) + ', found ndim=' +
    218                          str(ndim) + '. Full shape received: ' +
--> 219                          str(tuple(shape)))
    220     if spec.max_ndim is not None:
    221       ndim = x.shape.rank

ValueError: Input 0 of layer lstm is incompatible with the layer: expected ndim=3, found ndim=2. Full shape received: (10, 8)

固定代码：

LSTM 期望输入一个 3D 张量，形状为 [batch, timesteps, feature]。

import tensorflow as tf
import numpy as np

inputs = tf.random.normal([32, 10, 8])
simple_lstm = tf.keras.layers.LSTM(4)
output = simple_lstm(inputs) 

print(output)

输出：

tf.Tensor(
[[ 0.06065299  0.3409816   0.32014534  0.06411142]
 [ 0.00128129 -0.38577533 -0.11682745  0.10254318]
 [ 0.20580113 -0.14564514  0.11878442 -0.10241538]
 [-0.19621769  0.159462   -0.14459077 -0.06216513]
 [-0.10081916  0.17638563  0.07971784  0.1896367 ]
 [ 0.33309937  0.07162716 -0.08868891 -0.00883376]
 [ 0.17272277 -0.34112597  0.099504    0.0996887 ]
 [ 0.21682273  0.00900807  0.5081149   0.02028211]
 [ 0.2525146   0.04386558 -0.09498325  0.10461893]
 [ 0.21941815 -0.3566848  -0.05213086  0.18148176]
 [ 0.22719224 -0.29461107  0.07673237 -0.1128229 ]
 [ 0.00436124 -0.14181408  0.0085922  -0.49300092]
 [-0.15231487  0.06897711 -0.30905092  0.06128961]
 [ 0.25731358 -0.03430091 -0.2770667   0.14336488]
 [-0.09124507  0.12587348  0.04689778 -0.12776485]
 [ 0.15820538 -0.03368861  0.01726492 -0.02723333]
 [ 0.04661836 -0.06789393  0.0413182   0.14790519]
 [-0.04050795  0.18242228 -0.14569572  0.00631422]
 [ 0.12048664 -0.01098521 -0.19472744  0.15155892]
 [-0.20725198  0.1710444  -0.3829169   0.01446645]
 [ 0.06099659  0.15198827 -0.18342684  0.10251417]
 [ 0.01376235 -0.07642762  0.16639794  0.02458677]
 [ 0.21958975 -0.08766301 -0.02507084  0.00333961]
 [-0.15135197 -0.00785332 -0.02620712 -0.15565342]
 [ 0.07218299 -0.0798007   0.01710635 -0.2993008 ]
 [ 0.41292062 -0.2292722  -0.14371048  0.2036194 ]
 [ 0.19662695 -0.10295419 -0.01334361 -0.3022645 ]
 [-0.2587392  -0.10956616  0.10394819 -0.3500641 ]
 [-0.0293685  -0.25388837  0.07153057  0.02657588]
 [ 0.23911244 -0.3574759   0.06245361 -0.04481344]
 [-0.32070398  0.03763141  0.03036258 -0.2610327 ]
 [-0.13514674 -0.14885807  0.2496089  -0.12311874]], shape=(32, 4), dtype=float32)