Where to install zlib and cudnn files?

I’m new to using deep learning software. I do a lot of bio-imaging and image processing and I need to use software with deep learning capabilities. I installed the CUDA toolkit and now I need to install cuDNN. I downloaded the zlib and cuDNN zip files and the instructions don’t indicate into which system folders the contents of the zip files should go. Where do I put the dll_x64 and static_x64 zlib files and where do I put the cuDNN bin, include, and lib folders? Also, how do I add the directory path of zlibwapi.dll to the environment variable PATH? I have a GeForce RTX 3080 laptop GPU in an Alienware 17 R1 laptop with Windows 11. Thank you for your help.
Mary

2 Likes

Hi,

Please refer cuDNN installation guide to get more info on how to install zlib, cuDNN.

Thank you.

Hi, I’m having the same problem even after following the instructions in the installation guide while using Tensorflow.
To be more precise, I’ve tried the following things:

  1. Installing cuDNN 8.4.0 from the zip, by moving the files in the suggested “C:/Program Files/NVIDIA/…” and adding the path to the PATH environment variable (this results in the same error).
  2. Installing cuDNN 8.3.3.40 with the installer and adding the path to the PATH (this results in the same error).
  3. Copying the files into the CUDA folder, as suggested in some guides (if I do this, Tensorflow correctly identify the GPU but the program crashes after loading cuDNN without giving any error on python, just an error code and an error event on windows.

As suggested in the guide, I have installed zlib123dll and added the path on PATH.

I really don’t understand whether I’m doing something wrong or if there are some compatibility issues.

Below are listed the version I’m using:

  • CUDA 11.6
  • cuDNN 8.4.0 (and 8.3.3.40 as previously mentioned)
  • python 10
  • tensorflow nighly 2.10.0.20220420
  • zlib from NVIDIA guide link (zlib123dll.zip)
    OS: Windows 10
    Hardware: RTX 3090

Thanks in advance for helping

1 Like

Similar problem. The guide simply says:

“Add the directory path of zlibwapi.dll to the environment variable PATH.” Since Windows downloaded it to my Downloads folder, and I don’t want my Downloads folder in my Path, I copied it to its own directory off my c: drive and added that to my path. The guide doesn’t specify if I should include the ‘dll_x64’ folder or the parent of that folder, nor does it say if I should have a trailing backslash, but all 4 ways give the same error.

I still get:
“Could not locate zlibwapi.dll. Please make sure it is in your library path!”
trying to run a TensorFlow program with Python.

The instructions are wrong!

You need to copy zlibwapi.dll directly into:
C:\Program Files\NVIDIA\CUDNN\v 8.x\bin

Adding a path to it like the instructions say is not enough!

5 Likes

This still didnt work for me

it worked for me. Thanks bro

From here
Install zlib-wapi by conda (conda-forge)

conda install -c conda-forge zlib-wapi

This is an alternative way if one is using conda.

8 Likes

working fine.

Thank you bro, you saved me!