github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #1265] Correctly set up WSL environment to enable CUDA in building Ollama #62684

New Issue
Closed
opened 2026-05-03 09:57:57 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-05-03 09:57:57 -05:00
Owner
Copy Link

Originally created by @taweili on GitHub (Nov 24, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1265

After probing around the environment setup and the source codes for a few days, I finally figured out how to correctly build Ollama to support CUDA under WSL.

  1. WSL, by default, includes Windows's PATH, and there is an nvcc if one has installed the cuda environment in Windows.
  2. The default path to Linux's cuda isn't probably set in the environment

To fix this:

  1. Take out Windows path inclusion in the /etc/wsl.conf
[interop]
appendWindowsPath = false
  1. Set up Linux for CUDA development in your ~/.bashrc
# set up cuda
export PATH=/usr/local/cuda/bin:$PATH
export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/usr/local/cuda/lib64

And follow the build instructions to generate and build ollama

Wow, it's totally awesome to be able to use the GPU with Ollama. 60+ tokens/minute on a Titan RTX!

Originally created by @taweili on GitHub (Nov 24, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1265 After probing around the environment setup and the source codes for a few days, I finally figured out how to correctly build Ollama to support CUDA under WSL. 1. WSL, by default, includes Windows's PATH, and there is an nvcc if one has installed the cuda environment in Windows. 2. The default path to Linux's cuda isn't probably set in the environment To fix this: 1. Take out Windows path inclusion in the `/etc/wsl.conf` ``` [interop] appendWindowsPath = false ``` 2. Set up Linux for CUDA development in your `~/.bashrc` ``` # set up cuda export PATH=/usr/local/cuda/bin:$PATH export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/usr/local/cuda/lib64 ``` And follow the build instructions to generate and build ollama Wow, it's totally awesome to be able to use the GPU with Ollama. 60+ tokens/minute on a Titan RTX!
GiteaMirror commented 2026-05-03 09:57:58 -05:00
Author
Owner
Copy Link

@BruceMacD commented on GitHub (Nov 24, 2023):

Thanks for documenting this @taweili, for future reference the compiled released will also work with CUDA from WSL:
https://www.ollama.ai/download/linux

Closing this for now as there doesn't seem to be any pending actions, but it can be referenced by people searching in the future.

<!-- gh-comment-id:1826109185 --> @BruceMacD commented on GitHub (Nov 24, 2023): Thanks for documenting this @taweili, for future reference the compiled released will also work with CUDA from WSL: https://www.ollama.ai/download/linux Closing this for now as there doesn't seem to be any pending actions, but it can be referenced by people searching in the future.
GiteaMirror commented 2026-05-03 09:57:59 -05:00
Author
Owner
Copy Link

@taweili commented on GitHub (Nov 25, 2023):

@BruceMacD Thanks. I posted this mostly for the record in case it may be useful. I have been searching for solution on Ollama not using the GPU in WSL since 0.1.10 and updating to 0.1.11 didn't help. I decided to compile the codes myself and found that WSL's default path setup could be a problem.

I tried both releases and I can't find a consistent answer on whether or not looking at the issues posted here. Some said that it wasn't supposed to work and some had it working.

I also notice that my compiled binary is about 20% bigger than the released version. I don't yet have a chance to investigate the difference.

<!-- gh-comment-id:1826248622 --> @taweili commented on GitHub (Nov 25, 2023): @BruceMacD Thanks. I posted this mostly for the record in case it may be useful. I have been searching for solution on Ollama not using the GPU in WSL since 0.1.10 and updating to 0.1.11 didn't help. I decided to compile the codes myself and found that WSL's default path setup could be a problem. I tried both releases and I can't find a consistent answer on whether or not looking at the issues posted here. Some said that it wasn't supposed to work and some had it working. I also notice that my compiled binary is about 20% bigger than the released version. I don't yet have a chance to investigate the difference.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#62684
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?