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 一起使用时，我不断收到此错误，我已根据说明多次安装 CUDA、cuDNN 和所有驱动程序。但似乎没有任何效果。 如果我使用笔记本，那么 TensorFlow 使用 CPU，通过 VS 代码笔记本扩展，我可以使用 gpu，但是当我尝试将其作为普通 python 文件运行时，它会在第一个 epoch 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sn-p：

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)

Jupyter-notebook、VS code notebook 扩展和普通的 python 文件使用了相同的代码

设备规格：

处理器：英特尔 i5 显卡：英伟达 Geforce 1050ti

Cuda 版本：11.5 cuDNN 版本：8.3

Answer 1

对于那些仍然有这个问题的人，请确保你也完成了this step：

下载、解压缩并将zlibwapi.dll 添加到您的系统路径。

我为此浪费了半个小时，所以你不必也这样做。祝你好运！

Answer 2

和我一样的“错误”。即使我已经用“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 但将 cuda 版本保持为 11.5 时，“错误”消失了。 （也需要重新编译） 所以我认为这个错误必须在“cuDNN”上。

请看下一个答案。效果很好。