Failed to get convolution algorithm. This is probably because cuDNN failed to initialize

最新推荐文章于 2021-11-02 14:22:03 发布
原创 最新推荐文章于 2021-11-02 14:22:03 发布
· 458 阅读 · 0 ·
版权
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
文章标签：

#tensorflow

思考原因：显存不够用？但是下面的代码是限制显存的使用，能够使用的显存会更少，怎么就能运行程序了呢？可能是这样的：TensorFlow 假设显存是充足的，然后就计算需要的显存用量，结果发现显存实际大小不可满足需求，就报错。但是假如我们设定了显存用量，TensorFlow 就知道最多可以用多少显存，然后合理利用显存来训练模型。

解决办法：设置 Tensorflow 最多可使用 GPU 显存大小。把下面这段代码复制到你的代码最前面即可。

# GPU 按需分配
config = tf.compat.v1.ConfigProto(allow_soft_placement=True)
config.gpu_options.per_process_gpu_memory_fraction = 0.3
tf.compat.v1.keras.backend.set_session(tf.compat.v1.Session(config=config))
解决Failed to get convolution algorithm. This is probably because cuDNN failed to initialize
qq_38835585的博客
09-23 8422
问题如下： UnknownError (see above for traceback): 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. [[node conv2d_1/convolution (defined at C:\Us
【深度学习】Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so t
m0_52357468的博客
03-09 1161
本人TensorFlow-gpu 2.0 电脑显卡1650ti。 有时候电脑明明gpu没任何占用，跑深度学习代码还是会出现以下这个报错： UnknownError: 2 root error(s) found. (0) Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log messa
Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try…
01-06
This is probably because cuDNN failed to initialize, so try…”于是打开retrain.py文件，在前面加上如下代码： from tensorflow.compat.v1 import ConfigProto from tensorflow.compat.v1 import ...
Error : Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so
12-22
This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above. 主要原因是版本问题，还有可能是没有安装cuda / cudnn ,具体可在pycharm的设置中查看
UnknownError: Failed to get convolution algorithm. This is probably because cuDNN failed 报错 Jupyter
Welcome
11-02 499
未知错误：获取卷积算法失败。这可能是因为cuDNN失败了。 这可能是由于显存被占用，详细原因可看这篇。 关于keras or tensorflow 显存占用问题 ： keras or tensorflow 显存占用问题 - 简书 解决方法： 1. 如果用的是Keras，在jupyter notebook 最开始添加代码： import tensorflow as tf import numpy as np import keras config = tf.compat.v1.ConfigP
tensorflow卷积报错Failed to get convolution algorithm. This is probably because cuDNN failed t
集电极
07-05 5091
tensorflow2.0卷积报错Failed to get convolution algorithm. This is probably because cuDNN failed t 说明：环境是tensorflow2.0.0,测试GPU没有问题。 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 lookin
【ERROR】Failed to get convolution algorithm. This is probably because cuDNN failed to initialize
这个人很蓝
09-06 661
1、有可能是显存不足，尝试减小batchsize 2、os.environ['TF_FORCE_GPU_ALLOW_GROWTH'] = 'true' 这个语句的意思是，TensorFlow 在分配显存的时候，设置为可以逐步分配，需要多少显存，就一点点往上添，直到够用为止。比如GPU显存为8GB，开始训练一个模型，TensorFlow 先分配100MB，然后200MB 、300MB……最后在700MB处停住了，因为实际700MB就够用了。参考自 ...
“Failed to get convolution algorithm. This is probably because cuDNN failed to initialize”错误的解决办法
qq_40245400的博客
10-21 475
https://blog.csdn.net/tsyccnh/article/details/102938368 最近在使用TF2.0。运行程序出现以下错误。 “Failed to get convolution algorithm. This is probably because cuDNN failed to initialize” 一开始怀疑是CUDA和CuDNN配置错误（要求版本匹配）。反复试验后，还是有这个错误。 最后发现可能是GPU内存不足造成的。需要在程序前加以下一段代码： from ten
解决：Failed to get convolution algorithm. This is probably because cuDNN failed to initialize
wangxu4980347的博客
12-01 1325
解决报错 "Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so trylooking to see if a warning log message was printed above. [Op:Conv2D] " 代码块前面加上三行： physical_devic...
(0) unknown: 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.
04-11
this is probably because cudnn failed to initialize, so try looking to see if a warning log message was printed above.'." 的意思是什么？ A: 这个错误提示表示在获取卷积算法时出现了问题，原因很可能是...
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值