突变::应用错误：缺少节点“顺序/密集/MatMul”存在扇出“顺序/密集/BiasAdd”

Question

创建一个简单的线性回归模型来检测马力每加仑的英里数，由于某种原因，该模型根本无法拟合，我可以预测，总结等。 - 这不是整个代码，只是创建模型的部分。所有其他代码都在清理数据

# Building horsepower normalizer
horsepower_features = np.array(train_features['Horsepower'])
horsepower_val_features = np.array(val_features['Horsepower'])
horsepower_test = np.array(test_features['Horsepower'])

horsepower_labels = np.array(train_labels)
horsepower_val_labels = np.array(val_labels)
horsepower_test_labels = np.array(test_labels)

horsepower_normalizer = layers.Normalization(input_shape=[1,], axis=None)
horsepower_normalizer.adapt(horsepower_features)

print(horsepower_normalizer.mean.numpy())
print('Dataset not normalized', horsepower_features)
print('Dataset normalized', horsepower_normalizer(horsepower_features).numpy)

horsepower_model = keras.Sequential([
    horsepower_normalizer,
    layers.Dense(units=1)
])
horsepower_model.summary()

print(horsepower_model.predict(horsepower_features[:10]))

horsepower_model.compile(
    loss=losses.SparseCategoricalCrossentropy(from_logits=True),
    optimizer=optimizers.legacy.Adam(learning_rate=0.1),
    metrics=['accuracy'])

history = horsepower_model.fit(
    horsepower_features,
    horsepower_labels,
    epochs=100,
    verbose=1,
    validation_data=(horsepower_val_features, horsepower_val_labels)
)

这是它返回的错误

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 normalization_1 (Normaliza  (None, 1)                 3         
 tion)                                                           
                                                                 
 dense (Dense)               (None, 1)                 2         
                                                                 
=================================================================
Total params: 5 (24.00 Byte)
Trainable params: 2 (8.00 Byte)
Non-trainable params: 3 (16.00 Byte)
_________________________________________________________________
2023-09-26 09:39:56.650231: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:114] Plugin optimizer for device_type GPU is enabled.
1/1 [==============================] - 0s 42ms/step
[[ 0.01643855]
 [-0.34483466]
 [-0.45892093]
 [ 0.8720856 ]
 [-0.99132353]
 [-0.07863335]
 [-0.30680588]
 [-0.26877713]
 [-0.17370522]
 [-0.32582027]]
Epoch 1/100
2023-09-26 09:39:56.799326: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:114] Plugin optimizer for device_type GPU is enabled.
2023-09-26 09:39:56.807283: F metal_plugin/src/graph/remapper/remapper.cc:3827] Mutation::Apply error: fanout 'sequential/dense/BiasAdd' exist for missing node 'sequential/dense/MatMul'.

Process finished with exit code 134 (interrupted by signal 6: SIGABRT)

我花了几个小时尝试修复它，卸载，重新安装 anaconda 等，但无法修复它。请帮忙。

Answer 1

我假设您正在 macOS tensorflow-metal 上部署。如果是，则tensorflow-macos和tensorflow-metal之间的版本不兼容。 例如。这一天，最新的 TensorFlow 版本是 2.15。最新的tensorflow-metal版本是1.1.0，与tensorflow==2.14.0兼容。 总之，您应该降级您的依赖项。

https://pypi.org/project/tensorflow-metal/ 请参阅项目描述了解更多信息。