无法在仅 TensorFlow CPU 版本中加载模型
Posted
技术标签:
【中文标题】无法在仅 TensorFlow CPU 版本中加载模型【英文标题】：Unable to load model in Tensorflow CPU only version 【发布时间】：2021-05-24 17:15:54 【问题描述】：环境：
Tensorflow：2.3.0（仅限 CPU） Python：3.8.5 GPU: 0 操作系统：Ubuntu 20.04 LTS问题陈述：
我很抱歉问了另一个新手问题，但我正在尝试使用 Tensorflow （仅限 CPU 版本） 中的 load_model() 方法加载模型。
I tensorflow/core/platform/profile_utils/cpu_utils.cc:104] CPU Frequency: 1996330000 Hz
I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x7fc360269ab0 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): Host, Default Version
尝试：
我尝试设置环境变量link
import os
os.environ['CUDA_VISIBLE_DEVICES'] = '-1'
import tensorflow as tf
from keras.models import load_model
model = tf.keras.models.load_model('path/to/location/model.model')
或者
import os
os.environ['CUDA_VISIBLE_DEVICES'] = ''
import tensorflow as tf
from keras.models import load_model
model = tf.keras.models.load_model('path/to/location/model.model')
注意：请检查模型是否在
.model扩展中
Q1。有没有办法检查我在.model 扩展中的模型？
编辑：
根据@kosa 回答model.summary() 给我以下输出。
Model: "model"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
image (InputLayer) [(None, 45, 168, 1)] 0
__________________________________________________________________________________________________
conv2d (Conv2D) (None, 45, 168, 16) 160 image[0][0]
__________________________________________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 23, 84, 16) 0 conv2d[0][0]
__________________________________________________________________________________________________
conv2d_1 (Conv2D) (None, 23, 84, 32) 4640 max_pooling2d[0][0]
__________________________________________________________________________________________________
max_pooling2d_1 (MaxPooling2D) (None, 12, 42, 32) 0 conv2d_1[0][0]
__________________________________________________________________________________________________
conv2d_2 (Conv2D) (None, 12, 42, 32) 9248 max_pooling2d_1[0][0]
__________________________________________________________________________________________________
max_pooling2d_2 (MaxPooling2D) (None, 6, 21, 32) 0 conv2d_2[0][0]
__________________________________________________________________________________________________
batch_normalization_v1 (BatchNo (None, 6, 21, 32) 128 max_pooling2d_2[0][0]
__________________________________________________________________________________________________
flatten (Flatten) (None, 4032) 0 batch_normalization_v1[0][0]
__________________________________________________________________________________________________
dense (Dense) (None, 64) 258112 flatten[0][0]
__________________________________________________________________________________________________
dense_1 (Dense) (None, 64) 258112 flatten[0][0]
__________________________________________________________________________________________________
dense_2 (Dense) (None, 64) 258112 flatten[0][0]
__________________________________________________________________________________________________
dense_3 (Dense) (None, 64) 258112 flatten[0][0]
__________________________________________________________________________________________________
dense_4 (Dense) (None, 64) 258112 flatten[0][0]
__________________________________________________________________________________________________
dense_5 (Dense) (None, 64) 258112 flatten[0][0]
__________________________________________________________________________________________________
dropout (Dropout) (None, 64) 0 dense[0][0]
__________________________________________________________________________________________________
dropout_1 (Dropout) (None, 64) 0 dense_1[0][0]
__________________________________________________________________________________________________
dropout_2 (Dropout) (None, 64) 0 dense_2[0][0]
__________________________________________________________________________________________________
dropout_3 (Dropout) (None, 64) 0 dense_3[0][0]
__________________________________________________________________________________________________
dropout_4 (Dropout) (None, 64) 0 dense_4[0][0]
__________________________________________________________________________________________________
dropout_5 (Dropout) (None, 64) 0 dense_5[0][0]
__________________________________________________________________________________________________
char_1 (Dense) (None, 36) 2340 dropout[0][0]
__________________________________________________________________________________________________
char_2 (Dense) (None, 36) 2340 dropout_1[0][0]
__________________________________________________________________________________________________
char_3 (Dense) (None, 36) 2340 dropout_2[0][0]
__________________________________________________________________________________________________
char_4 (Dense) (None, 36) 2340 dropout_3[0][0]
__________________________________________________________________________________________________
char_5 (Dense) (None, 36) 2340 dropout_4[0][0]
__________________________________________________________________________________________________
char_6 (Dense) (None, 36) 2340 dropout_5[0][0]
==================================================================================================
Total params: 1,576,888
Trainable params: 1,576,824
Non-trainable params: 64
__________________________________________________________________________________________________
None
【问题讨论】：
是什么让你认为这没有成功？ @gobrewers14，我遇到了上述错误。 既然，我是新手，很高兴知道拒绝投票的原因。这样我就可以避免这样的错误。 【参考方案1】：可能没有任何错误。请尝试model.summary() 并检查其输出。
【讨论】：
看起来工作正常。为什么说它没有加载？以上是关于无法在仅 TensorFlow CPU 版本中加载模型的主要内容，如果未能解决你的问题，请参考以下文章
Spyder 无法在 conda 环境中加载 tensorflow，尽管 python 可以
无法在 tensorflow 官方 resnet 模型中加载用于 eval 的图像
无法在 Keras 2.1.0（使用 Tensorflow 1.3.0）中保存的 Keras 2.4.3（使用 Tensorflow 2.3.0）中加载 Keras 模型