Keras Sequential模型输入形状不兼容报错求助

问题代码 #

from torch.nn.modules import activation
from tensorflow.keras import models
from keras.models import Sequential 
from tensorflow.keras.layers import concatenate
from tensorflow.keras.utils import plot_model
from tensorflow.keras.layers import add
from keras.layers.core import Dense, Dropout, Activation
from tensorflow.keras.layers import Input

from torch import relu

input_shape = en_X_train.shape[1]
tf.keras.backend.clear_session()


input_shape = en_X_train.shape

modeloA = tf.keras.models.Sequential()

modeloA.add(keras.layers.Input(shape=input_shape))
modeloA.add(keras.layers.BatchNormalization())
modeloA.add(keras.layers.Dense(256, activation='relu'))
modeloA.add(keras.layers.Dense(256, activation='relu'))
modeloA.add(keras.layers.Dense(128, activation='linear'))
modeloA.add(keras.layers.BatchNormalization())
modeloA.add(keras.layers.Activation('relu'))
modeloA.add(keras.layers.Dropout(0.25))
modeloA.add(keras.layers.Dense(1, activation='sigmoid'))

modeloA.summary()
modeloA.layers[2].trainable = False
modeloA.compile(loss=tf.keras.losses.BinaryCrossentropy(),
            optimizer=keras.optimizers.Adam(learning_rate=0.0001, beta_1=-0.075),
            metrics=[tf.keras.metrics.Precision(), tf.keras.metrics.Recall()])
HistA = modeloA.fit(en_X_train, en_y_train, batch_size=16, epochs=25, verbose=1)

模型摘要 #

Model: "sequential"
----------------------------------------------------------------
Layer (type)                Output Shape              Param #
================================================================
batch_normalization (BatchNormalization)  (None, 42040, 512)       2048
dense (Dense)               (None, 42040, 256)        131328
dense_1 (Dense)             (None, 42040, 256)        65792
dense_2 (Dense)             (None, 42040, 128)        32896
batch_normalization_1 (BatchNormalization)  (None, 42040, 128)       512
activation (Activation)     (None, 42040, 128)        0
dropout (Dropout)           (None, 42040, 128)        0
dense_3 (Dense)             (None, 42040, 1)          129
================================================================
Total params: 232,705
Trainable params: 231,425
Non-trainable params: 1,280

报错信息 #

Epoch 1/25
ValueError                                Traceback (most recent call last)
 in 
33                         optimizer=keras.optimizers.Adam(learning_rate=0.0001, beta_1=-0.075),
34             metrics=[tf.keras.metrics.Precision(), tf.keras.metrics.Recall()])
---> 35 HistA = modeloA.fit(en_X_train, en_y_train, batch_size=16, epochs=25, verbose=1)

1 frames
/usr/local/lib/python3.8/dist-packages/keras/engine/training.py in tf__train_function(iterator)
13                 try:
14                     do_return = True
---> 15                     retval_ = ag__.converted_call(ag__.ld(step_function), (ag__.ld(self), ag__.ld(iterator)), None, fscope)
16                 except:
17                     do_return = False

ValueError: in user code:

File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 1051, in train_function  *
    return step_function(self, iterator)
File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 1040, in step_function  **
    outputs = model.distribute_strategy.run(run_step, args=(data,))
File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 1030, in run_step  **
    outputs = model.train_step(data)
File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 889, in train_step
    y_pred = self(x, training=True)
File "/usr/local/lib/python3.8/dist-packages/keras/utils/traceback_utils.py", line 67, in error_handler
    raise e.with_traceback(filtered_tb) from None
File "/usr/local/lib/python3.8/dist-packages/keras/engine/input_spec.py", line 264, in assert_input_compatibility
    raise ValueError(f'Input {input_index} of layer "{layer_name}" is '

ValueError: Input 0 of layer "sequential" is incompatible with the layer: expected shape=(None, 42040, 512), found shape=(None, 512)

解决方法 #

核心问题分析 #

报错直接原因是输入形状不匹配：模型期望输入形状为(None, 42040, 512)，但实际输入是(None, 512)。这是因为代码错误地将包含样本数的完整数据集形状传给了Input层，而Keras的Input层仅需指定特征维度，样本维度由框架自动用None表示。

另外存在两个次要问题：

• Adam优化器的beta_1参数设为负数，该参数必须在0-1范围内，否则会导致优化器异常。
• 导入了未使用的PyTorch模块，属于冗余导入，易造成混淆。

修正后的代码 #

from tensorflow.keras import models
from tensorflow.keras.models import Sequential 
from tensorflow.keras.layers import Dense, Dropout, Activation, BatchNormalization, Input
from tensorflow.keras.losses import BinaryCrossentropy
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.metrics import Precision, Recall

# 仅保留特征维度作为输入形状
input_shape = en_X_train.shape[1]
tf.keras.backend.clear_session()

modeloA = Sequential()
# Input层仅传入特征维度，样本维度由Keras自动处理
modeloA.add(Input(shape=(input_shape,)))
modeloA.add(BatchNormalization())
modeloA.add(Dense(256, activation='relu'))
modeloA.add(Dense(256, activation='relu'))
modeloA.add(Dense(128, activation='linear'))
modeloA.add(BatchNormalization())
modeloA.add(Activation('relu'))
modeloA.add(Dropout(0.25))
modeloA.add(Dense(1, activation='sigmoid'))

modeloA.summary()
modeloA.layers[2].trainable = False
# 修正beta_1为合理值0.9
modeloA.compile(loss=BinaryCrossentropy(),
                optimizer=Adam(learning_rate=0.0001, beta_1=0.9),
                metrics=[Precision(), Recall()])
HistA = modeloA.fit(en_X_train, en_y_train, batch_size=16, epochs=25, verbose=1)

关键修正点 #

1. 输入形状修正：将input_shape = en_X_train.shape改为input_shape = en_X_train.shape[1]，并在Input层传入(input_shape,)，确保模型期望输入形状与实际训练数据匹配。
2. 优化器参数修正：把beta_1=-0.075改为beta_1=0.9（Adam默认值之一，符合参数范围要求）。
3. 清理冗余导入：删除未使用的PyTorch模块和Keras冗余导入，统一使用tensorflow.keras下的模块，避免命名冲突。

内容的提问来源于stack exchange，提问作者Carlos Eduardo