无法加载库 cudnn_cnn_infer64_8.dll。错误代码 126

Question

Could not load library cudnn_cnn_infer64_8.dll. Error code 126
Please make sure cudnn_cnn_infer64_8.dll is in your library path!

当我尝试将 TensorFlow 与 GPU 结合使用时，我不断收到此错误，我已根据说明多次安装了 CUDA、cuDNN 和所有驱动程序。但似乎没有任何作用。 如果我使用笔记本，那么 TensorFlow 使用 CPU，通过 VS code 笔记本扩展，我可以使用 GPU，但当我尝试将其作为普通 python 文件运行时，它会在第一个纪元停止会话。出现上述错误。

完整的终端输出：

Found 14630 validated image filenames belonging to 3 classes.
Found 1500 validated image filenames belonging to 3 classes.
2021-11-08 11:03:58.000354: I tensorflow/core/platform/cpu_feature_guard.cc:151] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX AVX2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2021-11-08 11:03:58.603592: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1525] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 2775 MB memory:  -> device: 0, name: NVIDIA GeForce GTX 1050 Ti, pci bus id: 0000:01:00.0, compute capability: 6.1        
Epoch 1/10
2021-11-08 11:04:07.306011: I tensorflow/stream_executor/cuda/cuda_dnn.cc:366] Loaded cuDNN version 8300
Could not load library cudnn_cnn_infer64_8.dll. Error code 126
Please make sure cudnn_cnn_infer64_8.dll is in your library path!
E:\MyWorkSpace\animal_detect>

代码片段：

import tensorflow as tf 
from tensorflow.keras.preprocessing.image import ImageDataGenerator 
from tensorflow.keras import layers 
from tensorflow.keras import Model 
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.applications.vgg16 import VGG16
import pandas as pd
import numpy as np

train_df = pd.read_csv('train.csv')
test_df = pd.read_csv('test.csv')
train_gen = ImageDataGenerator(rescale = 1./255.,rotation_range = 40, width_shift_range = 0.2, height_shift_range = 0.2, shear_range = 0.2, zoom_range = 0.2, horizontal_flip = True)
test_gen = ImageDataGenerator( rescale = 1.0/255. )
train_set = train_gen.flow_from_dataframe(train_df,x_col='loc',y_col='label',batch_size=20,target_size=(224,224))
test_set = train_gen.flow_from_dataframe(test_df,x_col='loc',y_col='label',batch_size=20,target_size=(224,224))
base_model = VGG16(input_shape = (224, 224, 3),
include_top = False,
weights = 'imagenet')
for layer in base_model.layers:
    layer.trainable = False
x = layers.Flatten()(base_model.output)
x = layers.Dense(512, activation='relu')(x)
x = layers.Dropout(0.5)(x)
x = layers.Dense(3, activation='sigmoid')(x)

model = tf.keras.models.Model(base_model.input, x)

model.compile(optimizer = tf.keras.optimizers.RMSprop(learning_rate=0.0001), loss = 'categorical_crossentropy',metrics = ['acc'])
vgghist = model.fit(train_set, validation_data = test_set, steps_per_epoch = 100, epochs = 10)

相同的代码已用于 Jupyter-notebook、VS code 笔记本扩展以及普通的 python 文件

设备规格：

处理器：英特尔i5 GPU：Nvidia Geforce 1050ti

Cuda版本：11.5 cuDNN 版本：8.3

Answer 1

对于那些仍然遇到此问题的人，请确保您也已完成此步骤：

下载、解压并将

zlibwapi.dll

添加到您的系统路径。祝你好运！

Answer 2

和我一样的“错误”。尽管我已经用“Cuda版本：11.5 cuDNN版本：8.3”重新编译了tensorflow-gpu 2.6.0。 当我将 cudnn 版本更改为 8.2 但保留 cuda 版本为 11.5 时，“错误”消失了。 （也需要重新编译） 所以我认为这个错误一定是在“cuDNN”上。

请参阅androidu的回答。效果非常好。