Mirrored_Repos/InvokeAI
1
0
Fork 0
mirror of https://github.com/invoke-ai/InvokeAI synced 2024-08-30 20:32:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

Added extra steps to update the Cudnnn DLL found in the Torch packages

Browse Source 
I added extra steps to update the Cudnnn DLL found in the Torch package because it wasn't optimised or didn't use the lastest version. So manually updating it can speed up iteration but the result might differ from each card. Exemple i passed from 3 it/s to a steady 20 it/s.
This commit is contained in:
Keerigan45 2023-09-05 06:54:06 +02:00 committed by GitHub
parent 7a30162583
commit 04c0a83bff
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 25 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
docs/installation/030_INSTALL_CUDA_AND_ROCM.md
View File

@ -57,6 +57,31 @@ familiar with containerization technologies such as Docker.
For downloads and instructions, visit the [NVIDIA CUDA Container
Runtime Site](https://developer.nvidia.com/nvidia-container-runtime)
### Cudnnn Installation*
1) Find the InvokeAI folder
2) click on .venv folder should look like that YourInvokeFolderHere\.venv
3) Click on Lib folder should look like that YourInvokeFolderHere\.venv\Lib
4) Click on site-packages folder should look like YourInvokeFolderHere\.venv\Lib\site-packages
5) Find Torch directory when finded click on it should look like that YourInvokeFolderHere\InvokeAI\.venv\Lib\site-packages\torch
6) Find the lib folder should look like that YourInvokeFolderHere\.venv\Lib\site-packages\torch\lib
7) __Copy everything inside the folder as a Backup in whatever folder you want, it's just in case.__
8) Go to https://developer.nvidia.com/cudnn
9) Log-in Or Create an account if you're not already connected
10) Download the latest version
11) Go to the folder and extract it.
12) Find the bin folder E\cudnn-windows-x86_64-__Whatever Version__\bin
13) Copy the 7 dll
14) Go Back to YourInvokeFolderHere\.venv\Lib\site-packages\torch\lib
15) Paste the 7 dll took earlier. It should ask for replacement, accept it.
16) Enjoy !
__Very Important: You should Copy everything inside the folder of the torch lib. You do not moove It.__
*Note:
If _no change is seen or bug appear__ follow the same step instead just copy the Torch/lib back up folder you made earlier and replace it !
This optimization is normally intented for the newer version of graphics card (4th series 3th series) but results have been seen with older graphics card.
So giving a try could be good.
### Torch Installation
When installing torch and torchvision manually with `pip`, remember to provide