github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 16:40:08 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #3624] [v0.1.32-pre for Windows] Can't use ollama run with OLLAMA_MODELS env set #48745

New Issue
Closed
opened 2026-04-28 09:11:17 -05:00 by GiteaMirror · 14 comments
GiteaMirror commented 2026-04-28 09:11:17 -05:00
Owner
Copy Link

Originally created by @mann1x on GitHub (Apr 13, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/3624

Originally assigned to: @bmizerany on GitHub.

What is the issue?

ollama run can't be used anymore with OLLAMA_MODELS environment set
instead:
Error: OLLAMA_MODELS must only be set for 'ollama serve'

What did you expect to see?

The ollama chat prompt

Steps to reproduce

set OLLAMA_MODELS system environment variable
type ollama run anymodel

Are there any recent changes that introduced the issue?

The latest prerelease

OS

Windows

Architecture

amd64

Platform

No response

Ollama version

0.1.32-pre

GPU

No response

GPU info

No response

CPU

AMD

Other software

No response

Originally created by @mann1x on GitHub (Apr 13, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/3624 Originally assigned to: @bmizerany on GitHub. ### What is the issue? ollama run can't be used anymore with OLLAMA_MODELS environment set instead: `Error: OLLAMA_MODELS must only be set for 'ollama serve'` ### What did you expect to see? The ollama chat prompt ### Steps to reproduce set OLLAMA_MODELS system environment variable type ollama run anymodel ### Are there any recent changes that introduced the issue? The latest prerelease ### OS Windows ### Architecture amd64 ### Platform _No response_ ### Ollama version 0.1.32-pre ### GPU _No response_ ### GPU info _No response_ ### CPU AMD ### Other software _No response_
GiteaMirror added the bug label 2026-04-28 09:11:17 -05:00
GiteaMirror commented 2026-04-28 09:11:19 -05:00
Author
Owner
Copy Link

@MoonRide303 commented on GitHub (Apr 13, 2024):

Same problem for me - only change was updating from 0.1.31 to 0.1.32-rc1, and now ollama run [model_name] results with Error: OLLAMA_MODELS must only be set for 'ollama serve' (for any model). Windows, too.

<!-- gh-comment-id:2053698269 --> @MoonRide303 commented on GitHub (Apr 13, 2024): Same problem for me - only change was updating from 0.1.31 to 0.1.32-rc1, and now `ollama run [model_name]` results with `Error: OLLAMA_MODELS must only be set for 'ollama serve'` (for any model). Windows, too.
GiteaMirror commented 2026-04-28 09:11:20 -05:00
Author
Owner
Copy Link

@Real00 commented on GitHub (Apr 15, 2024):

Maybe it's related to this PR
https://github.com/ollama/ollama/pull/3470

<!-- gh-comment-id:2055580806 --> @Real00 commented on GitHub (Apr 15, 2024): Maybe it's related to this PR https://github.com/ollama/ollama/pull/3470
GiteaMirror commented 2026-04-28 09:11:21 -05:00
Author
Owner
Copy Link

@AjmalShajahan commented on GitHub (Apr 15, 2024):

It was added in #3470
You can try this in powershell

$env:OLLAMA_MODELS=""; ollama run MODEL

OLLAMA_MODELS is set to "" for current session, when downloading/updating models in same session you'll need to set it to the path where the models are kept.

<!-- gh-comment-id:2057065996 --> @AjmalShajahan commented on GitHub (Apr 15, 2024): It was added in #3470 You can try this in powershell `$env:OLLAMA_MODELS=""; ollama run MODEL` OLLAMA_MODELS is set to "" for current session, when downloading/updating models in same session you'll need to set it to the path where the models are kept.
GiteaMirror commented 2026-04-28 09:11:22 -05:00
Author
Owner
Copy Link

@MoonRide303 commented on GitHub (Apr 15, 2024):

@AjmalShajahan but OLLAMA_MODELS points all the time to the path where I keep all my Ollama model files. I didn't change anything specifically for launching 0.1.32-rc1. 0.1.31 works fine, and 0.1.32-rc1 - just doesn't.

<!-- gh-comment-id:2057399782 --> @MoonRide303 commented on GitHub (Apr 15, 2024): @AjmalShajahan but OLLAMA_MODELS points all the time to the path where I keep all my Ollama model files. I didn't change anything specifically for launching 0.1.32-rc1. 0.1.31 works fine, and 0.1.32-rc1 - just doesn't.
GiteaMirror commented 2026-04-28 09:11:23 -05:00
Author
Owner
Copy Link

@MoonRide303 commented on GitHub (Apr 15, 2024):

@jmorganca @bmizerany Could you guys re-think that change (https://github.com/ollama/ollama/pull/3470)? I don't get it how that "helpful error message" solves anything - it just breaks essential functionality like "ollama run" for me.

<!-- gh-comment-id:2057420393 --> @MoonRide303 commented on GitHub (Apr 15, 2024): @jmorganca @bmizerany Could you guys re-think that change (https://github.com/ollama/ollama/pull/3470)? I don't get it how that "helpful error message" solves anything - it just breaks essential functionality like "ollama run" for me.
GiteaMirror commented 2026-04-28 09:11:23 -05:00
Author
Owner
Copy Link

@mann1x commented on GitHub (Apr 15, 2024):

OLLAMA_MODELS is set to "" for current session, when downloading/updating models in same session you'll need to set it to the path where the models are kept.

Yes it works of course but I run ollama on Windows, the OLLAMA_MODELS is set as system env variable. Can't be otherwise.

I don't really understand what was the problem with run on that PR but it breaks completely the client functionality while it looks the aim was to only show a diagnostic message.
Better to find a different solution, will try to understand what was the original issue.

If such a diagnostic message is needed maybe it's better to have it print out only when OLLAMA_DEBUG is set.

<!-- gh-comment-id:2057481157 --> @mann1x commented on GitHub (Apr 15, 2024): > OLLAMA_MODELS is set to "" for current session, when downloading/updating models in same session you'll need to set it to the path where the models are kept. Yes it works of course but I run ollama on Windows, the OLLAMA_MODELS is set as system env variable. Can't be otherwise. I don't really understand what was the problem with `run` on that PR but it breaks completely the client functionality while it looks the aim was to only show a diagnostic message. Better to find a different solution, will try to understand what was the original issue. If such a diagnostic message is needed maybe it's better to have it print out only when `OLLAMA_DEBUG` is set.
GiteaMirror commented 2026-04-28 09:11:24 -05:00
Author
Owner
Copy Link

@mann1x commented on GitHub (Apr 15, 2024):

I had a look at the code but I couldn't see any reason why run should care about the OLLAMA_MODELS env...

<!-- gh-comment-id:2057526802 --> @mann1x commented on GitHub (Apr 15, 2024): I had a look at the code but I couldn't see any reason why `run` should care about the `OLLAMA_MODELS` env...
GiteaMirror commented 2026-04-28 09:11:25 -05:00
Author
Owner
Copy Link

@MoonRide303 commented on GitHub (Apr 15, 2024):

@mann1x run uses it to know the path where model files are stored - at least that how it works in 0.1.31.

<!-- gh-comment-id:2057535554 --> @MoonRide303 commented on GitHub (Apr 15, 2024): @mann1x run uses it to know the path where model files are stored - at least that how it works in 0.1.31.
GiteaMirror commented 2026-04-28 09:11:25 -05:00
Author
Owner
Copy Link

@mann1x commented on GitHub (Apr 15, 2024):

@MoonRide303

run launches the client which only need OLLAMA_HOST; it's only connecting to the server, it's not interacting directly with the models.
Otherwise it wouldn't need it unset but set.
There must be another reason but I can't see it.

<!-- gh-comment-id:2057543612 --> @mann1x commented on GitHub (Apr 15, 2024): @MoonRide303 `run` launches the client which only need `OLLAMA_HOST`; it's only connecting to the server, it's not interacting directly with the models. Otherwise it wouldn't need it unset but set. There must be another reason but I can't see it.
GiteaMirror commented 2026-04-28 09:11:26 -05:00
Author
Owner
Copy Link

@mann1x commented on GitHub (Apr 15, 2024):

We need @bmizerany input on what was his issue with OLLAMA_MODELS and run but I suspect this is a bug.
Anyway it's a no go for Windows, if there's an issue we must find a different solution.

<!-- gh-comment-id:2057561792 --> @mann1x commented on GitHub (Apr 15, 2024): We need @bmizerany input on what was his issue with `OLLAMA_MODELS` and `run` but I suspect this is a bug. Anyway it's a no go for Windows, if there's an issue we must find a different solution.
GiteaMirror commented 2026-04-28 09:11:27 -05:00
Author
Owner
Copy Link

@MoonRide303 commented on GitHub (Apr 15, 2024):

@mann1x I just use "ollama run" like "main" from llama.cpp. I get that it might be launching server in the background, but I don't set OLLAMA_HOST or launch server explicitely.

<!-- gh-comment-id:2057591529 --> @MoonRide303 commented on GitHub (Apr 15, 2024): @mann1x I just use "ollama run" like "main" from llama.cpp. I get that it might be launching server in the background, but I don't set OLLAMA_HOST or launch server explicitely.
GiteaMirror commented 2026-04-28 09:11:27 -05:00
Author
Owner
Copy Link

@mann1x commented on GitHub (Apr 15, 2024):

@MoonRide303 Yes that's right; OLLAMA_HOST doesn't need to be set, it defaults to localhost, and the server is launched in the background if it's not started already

<!-- gh-comment-id:2057596115 --> @mann1x commented on GitHub (Apr 15, 2024): @MoonRide303 Yes that's right; `OLLAMA_HOST` doesn't need to be set, it defaults to localhost, and the server is launched in the background if it's not started already
GiteaMirror commented 2026-04-28 09:11:29 -05:00
Author
Owner
Copy Link

@jmorganca commented on GitHub (Apr 16, 2024):

Hi folks, this should be fixed now, and sorry for the issue!

<!-- gh-comment-id:2058068834 --> @jmorganca commented on GitHub (Apr 16, 2024): Hi folks, this should be fixed now, and sorry for the issue!
GiteaMirror commented 2026-04-28 09:11:30 -05:00
Author
Owner
Copy Link

@mann1x commented on GitHub (Apr 16, 2024):

@jmorganca Thanks, we were just confused!

<!-- gh-comment-id:2058168630 --> @mann1x commented on GitHub (Apr 16, 2024): @jmorganca Thanks, we were just confused!
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 bug
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#48745
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?