Python crashes after cudnn update

AI & Data Science Deep Learning (Training & Inference) cuDNN
1

Hi,

after updating cudnn I get the folowing output:

2021-11-11 17:04:58.146514: I tensorflow/stream_executor/cuda/cuda_dnn.cc:366] Loaded cuDNN version 8300

Process finished with exit code -1073740791 (0xC0000409)

any reason why?

I have tensorflow of 2.7, cuda 11.5

thanks

2 Likes
2

Hi @user43227
Can you pls share with us the detailed logs?

Thanks!

3

Same issue. My tensorflow is 2.6, cuda 11.5

4

Same issue.

5

i was troubled by this issue for 3 days and I have tried many ways to solve it but nothing worked.So I have to downgraded my python from 3.9 to 3.6,Cuda from 11.5 to 11.2,tensorflow from 2.6 to 2.5 , and cudnn from a newer version(i forgot it lol) to 8.2.After that it finally worked.

1 Like
6

I also downgraded cudnn from 11.5 to 11.4 and it works fine. Using tensorflow 2.6

1 Like
7

Thanks. I got the same issue. Downgraded as per your suggestion and it worked.

8

0xC0000409 error → install 11.4 cudnn

1 Like
9

Hi @user43227 can you check if you have zlib available in your environment?
This was a new dependency introduced first in cuDNN 8.3.0. See the following guide for getting zlib

3 Likes
10

After I got the problem, I installed exactly 2.8 Tensorflow, exactly 11.2 cuda and exactly 8.1 cuDNN, as it’s written here here: Build from source on Windows  |  TensorFlow
I guess, the main problem is with cuDNN, because with simple forward neural networks everything works, but when I added Convolutional layers, I got the problem.

Right now I reinstalled, as I wrote below, and everything works now

11

Follow the instruction to install zlib then the problem can be solved. I just tried with the cudnn 8.4.

12

Thanks for confirming installing zlib worked for you,

13

Thanks for sharing its working for you

14

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.