tensorflow卷积报错Failed to get convolution algorithm. This is probably because cuDNN failed t

说明：环境是tensorflow,测试GPU没有问题。

在tensorflow2.0和tensorflow1.15我都有遇到这个错误，主要是显卡内存分配问题。

但在运行模型出现错误。

UnknownError: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above. [Op:Conv2D]

解决方法添加下面两行

import osos.environ['CUDA_VISIBLE_DEVICES'] = '/gpu:0'  # 运行程序，都会占用gpu0全部资源# 多个GPU时，如果运行只使用了一个的话，可以设置为‘/gpu:0,1’等等


# 另一种写法
os.environ['CUDA_DEVICE_ORDER'] = 'PCI_BUS_ID' # 按照PCI_BUS_ID顺序从0开始排列GPU设备
os.environ['CUDA_VISIBLE_DEVICES'] = "0,1" #设置当前使用的GPU设备为0,1号两个设备,名称依次为'/gpu:0'、'/gpu:1'。
#[0,1]和[1,0]排列的设备是不同的，排在前面的设备优先级高，运行程序的时候会优先使用。

其他一些解答（我测试没有成功，但可能是我遇到的情况不一致）

1.矛头指向了是因为显存分配没搞好造成的

解决代码

import tensorflow as tf
config = tf.compat.v1.ConfigProto(allow_soft_placement=True)
config.gpu_options.per_process_gpu_memory_fraction = 0.8
tf.compat.v1.keras.backend.set_session(tf.compat.v1.Session(config=config))

来分析一下这段代码以及问题原因：
众所周知，keras以TensorFlow做后台时，TensorFlow会默认吃掉所有可用的显存（即便此时没有用到做计算，这些显存也会显示占用来待命）

第二行的0.3，意思是限制每个模型只能分配30%的显存，不允许吃掉所有。当然这是理论值，当实际开始计算时，如果需要，依然可以超出这个理论值。

2 也是说显卡内存分配问题

from tensorflow.compat.v1 import ConfigProto
from tensorflow.compat.v1 import InteractiveSessionconfig = ConfigProto()
config.gpu_options.allow_growth = True
session = InteractiveSession(config=config)

这个设置为按需分配

tensorflow卷积报错Failed to get convolution algorithm. This is probably because cuDNN failed t相关推荐

Failed to get convolution algorithm. This is probably because cuDNN failed to initialize
Failed to get convolution algorithm. This is probably because cuDNN failed to initialize 目前版本的Tensor ...
解决UnknownError: Failed to get convolution algorithm. This is probably because cuDNN failed to initia
解决UnknownError: Failed to get convolution algorithm. This is probably because cuDNN failed to initia ...
UnknownError: Failed to get convolution algorithm. This is probably because cuDNN failed to
UnknownError: Failed to get convolution algorithm. This is probably because cuDNN failed to initiali ...
【已解决】UnknownError: Failed to get convolution algorithm. This is probably because cuDNN failed to ini
jupyter运行有如下问题 UnknownError: Failed to get convolution algorithm. This is probably because cuDNN fai ...
Failed to get convolution algorithm.This is probably because cuDNN failed to initialize
报错原因: 内存分配出现了问题解决方案: 在代码前加上如下代码 import tensorflow as tf gpus = tf.config.list_physical_devices(&quo ...
[tf] Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initial
使用 tf1 进行训练时,出现这个问题,原因是 GPU 显存不足(当时有别的程序正在使用 GPU)
UnknownError: Failed to get convolution algorithm. This is probably because cuDNN failed to initial
遇到这个问题除了网上说的加代码以外还可以关闭其他占用显卡的程序,重启pycharm再次运行, 一般就解决问题.
Failed to get convolution algorithm. This is probably because cuDNN failed to initialize,
关于这个错误,这个帖子给出了完善的解释 https://stackoverflow.com/questions/53698035/failed-to-get-convolution-algorithm ...
tensorflow出现UnknownError: Failed to get convolution algorithm. This is probably because cuDNN faile
错误: tensorflow.python.framework.errors_impl.UnknownError: Failed to get convolution algorithm. This ...

tensorflow卷积报错Failed to get convolution algorithm. This is probably because cuDNN failed t

tensorflow卷积报错Failed to get convolution algorithm. This is probably because cuDNN failed t

解决方法添加下面两行

其他一些解答（我测试没有成功，但可能是我遇到的情况不一致）

1.矛头指向了是因为显存分配没搞好造成的

2 也是说显卡内存分配问题

tensorflow卷积报错Failed to get convolution algorithm. This is probably because cuDNN failed t相关推荐

最新文章

热门文章