github-starred/ollama-ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-04-27 07:57:18 -05:00
Code Issues 2.2k Packages Projects Releases 100 Wiki Activity

Problem witch loading models in vram #8567

New Issue
Closed
opened 2025-11-12 14:45:57 -06:00 by GiteaMirror · 23 comments
GiteaMirror commented 2025-11-12 14:45:57 -06:00
Owner
Copy Link

Originally created by @Savin1play140 on GitHub (Nov 2, 2025).

I have gtx 750 with 2GB of vram, but LLM model can't load, i try use qwen3 with 0.6b, on Windows gpu do discovered but model not loading, on Linux gpu don't discovered because i can't install CUDA 12.9 in system. what should I do?

OS

Windows and Linux

GPU

msi gtx 750 2GB

CPU

intel core i3 3210

Ollama version

v0.12.9

Originally created by @Savin1play140 on GitHub (Nov 2, 2025). I have gtx 750 with 2GB of vram, but LLM model can't load, i try use qwen3 with 0.6b, on Windows gpu do discovered but model not loading, on Linux gpu don't discovered because i can't install CUDA 12.9 in system. what should I do? ## OS Windows and Linux ## GPU msi gtx 750 2GB ## CPU intel core i3 3210 ## Ollama version v0.12.9
GiteaMirror commented 2025-11-12 14:45:58 -06:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 2, 2025):

GTX 750 is supported so should work in either OS. Server log may help with debugging.

@rick-github commented on GitHub (Nov 2, 2025): GTX 750 is supported so should work in either OS. [Server log](https://github.com/ollama/ollama/blob/main/docs/troubleshooting.mdx) may help with debugging.
GiteaMirror commented 2025-11-12 14:45:58 -06:00
Author
Owner
Copy Link

@Savin1play140 commented on GitHub (Nov 2, 2025):

GTX 750 is supported so should work in either OS. Server log may help with debugging.

i know about support a gtx 750, how i can enable debug? preferably without rebuilding the Ollama

@Savin1play140 commented on GitHub (Nov 2, 2025): > GTX 750 is supported so should work in either OS. [Server log](https://github.com/ollama/ollama/blob/main/docs/troubleshooting.mdx) may help with debugging. i know about support a gtx 750, how i can enable debug? preferably without rebuilding the Ollama
GiteaMirror commented 2025-11-12 14:45:58 -06:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 2, 2025):

Set OLLAMA_DEBUG=2 in the server environment.

@rick-github commented on GitHub (Nov 2, 2025): Set `OLLAMA_DEBUG=2` in the [server environment](https://github.com/ollama/ollama/blob/main/docs/faq.mdx#how-do-i-configure-ollama-server).
GiteaMirror commented 2025-11-12 14:45:58 -06:00
Author
Owner
Copy Link

@Savin1play140 commented on GitHub (Nov 2, 2025):

GTX 750 is supported so should work in either OS. Server log may help with debugging.

And more, if i install cuda 12.9 in Linux (bypassing the default installation of CUDA 13), will this work?

@Savin1play140 commented on GitHub (Nov 2, 2025): > GTX 750 is supported so should work in either OS. [Server log](https://github.com/ollama/ollama/blob/main/docs/troubleshooting.mdx) may help with debugging. And more, if i install cuda 12.9 in Linux (bypassing the default installation of CUDA 13), will this work?
GiteaMirror commented 2025-11-12 14:45:58 -06:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 2, 2025):

Ollama should work with CUDA 13. Support was only added recently though, so if there are issues, the server log would be helpful.

@rick-github commented on GitHub (Nov 2, 2025): Ollama should work with CUDA 13. Support was only added recently though, so if there are issues, the server log would be helpful.
GiteaMirror commented 2025-11-12 14:45:58 -06:00
Author
Owner
Copy Link

@Savin1play140 commented on GitHub (Nov 2, 2025):

Ollama should work with CUDA 13. Support was only added recently though, so if there are issues, the server log would be helpful.

That's not the problem, I think the CUDA 13 won't work with my gtx 750.

@Savin1play140 commented on GitHub (Nov 2, 2025): > Ollama should work with CUDA 13. Support was only added recently though, so if there are issues, the server log would be helpful. That's not the problem, I think the CUDA 13 won't work with my gtx 750.
GiteaMirror commented 2025-11-12 14:45:59 -06:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 2, 2025):

You are correct, CUDA 13 supports CC 7.5 and above, GTX 750 is CC 5.0 and so requires CUDA 12.

@rick-github commented on GitHub (Nov 2, 2025): You are correct, CUDA 13 supports CC 7.5 and above, GTX 750 is CC 5.0 and so requires CUDA 12.
GiteaMirror commented 2025-11-12 14:45:59 -06:00
Author
Owner
Copy Link

@Savin1play140 commented on GitHub (Nov 2, 2025):

You are correct, CUDA 13 supports CC 7.5 and above, GTX 750 is CC 5.0 and so requires CUDA 12.

Will ollama work correctly if I replace CUDA 13, which installs Linux by default, with CUDA 12.9?

@Savin1play140 commented on GitHub (Nov 2, 2025): > You are correct, CUDA 13 supports CC 7.5 and above, GTX 750 is CC 5.0 and so requires CUDA 12. Will ollama work correctly if I replace CUDA 13, which installs Linux by default, with CUDA 12.9?
GiteaMirror commented 2025-11-12 14:45:59 -06:00
Author
Owner
Copy Link

@Savin1play140 commented on GitHub (Nov 2, 2025):

You are correct, CUDA 13 supports CC 7.5 and above, GTX 750 is CC 5.0 and so requires CUDA 12.
Will ollama work correctly if I replace CUDA 13, which installs Linux by default, with CUDA 12.9?

@Savin1play140 commented on GitHub (Nov 2, 2025): > You are correct, CUDA 13 supports CC 7.5 and above, GTX 750 is CC 5.0 and so requires CUDA 12. Will ollama work correctly if I replace CUDA 13, which installs Linux by default, with CUDA 12.9?
GiteaMirror commented 2025-11-12 14:46:00 -06:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 2, 2025):

Yes.

@rick-github commented on GitHub (Nov 2, 2025): Yes.
GiteaMirror commented 2025-11-12 14:46:00 -06:00
Author
Owner
Copy Link

@Savin1play140 commented on GitHub (Nov 2, 2025):

Yes.

Okay, thanks you.

@Savin1play140 commented on GitHub (Nov 2, 2025): > Yes. Okay, thanks you.
GiteaMirror commented 2025-11-12 14:46:00 -06:00
Author
Owner
Copy Link

@pdevine commented on GitHub (Nov 4, 2025):

I think this is answered? I'm going to go ahead and close the issue, but can reopen if it's still a problem.

@pdevine commented on GitHub (Nov 4, 2025): I think this is answered? I'm going to go ahead and close the issue, but can reopen if it's still a problem.
GiteaMirror commented 2025-11-12 14:46:00 -06:00
Author
Owner
Copy Link

@Savin1play140 commented on GitHub (Nov 4, 2025):

server.log
In addition, ollama starts loading the processor by ~20-30% after a request to load a model

@Savin1play140 commented on GitHub (Nov 4, 2025): [server.log](https://github.com/user-attachments/files/23333737/server.log) In addition, ollama starts loading the processor by ~20-30% after a request to load a model
GiteaMirror commented 2025-11-12 14:46:01 -06:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 4, 2025):

The log doesn't show a model load. ollama does see the GPU:

time=2025-11-04T12:14:55.245+03:00 level=INFO source=types.go:42 msg="inference compute"
 id=GPU-45737e26-eef3-de24-eb43-691dc0d0f26b filtered_id="" library=CUDA compute=5.0 name=CUDA0
 description="NVIDIA GeForce GTX 750" libdirs=ollama,cuda_v12 driver=12.9 pci_id=0000:01:00.0 type=discrete
 total="2.0 GiB" available="1.8 GiB"
@rick-github commented on GitHub (Nov 4, 2025): The log doesn't show a model load. ollama does see the GPU: ``` time=2025-11-04T12:14:55.245+03:00 level=INFO source=types.go:42 msg="inference compute" id=GPU-45737e26-eef3-de24-eb43-691dc0d0f26b filtered_id="" library=CUDA compute=5.0 name=CUDA0 description="NVIDIA GeForce GTX 750" libdirs=ollama,cuda_v12 driver=12.9 pci_id=0000:01:00.0 type=discrete total="2.0 GiB" available="1.8 GiB" ```
GiteaMirror commented 2025-11-12 14:46:01 -06:00
Author
Owner
Copy Link

@Savin1play140 commented on GitHub (Nov 5, 2025):

The log doesn't show a model load. ollama does see the GPU:

time=2025-11-04T12:14:55.245+03:00 level=INFO source=types.go:42 msg="inference compute"
 id=GPU-45737e26-eef3-de24-eb43-691dc0d0f26b filtered_id="" library=CUDA compute=5.0 name=CUDA0
 description="NVIDIA GeForce GTX 750" libdirs=ollama,cuda_v12 driver=12.9 pci_id=0000:01:00.0 type=discrete
 total="2.0 GiB" available="1.8 GiB"

yes, that's right, but the problem still exists

@Savin1play140 commented on GitHub (Nov 5, 2025): > The log doesn't show a model load. ollama does see the GPU: > ``` > time=2025-11-04T12:14:55.245+03:00 level=INFO source=types.go:42 msg="inference compute" > id=GPU-45737e26-eef3-de24-eb43-691dc0d0f26b filtered_id="" library=CUDA compute=5.0 name=CUDA0 > description="NVIDIA GeForce GTX 750" libdirs=ollama,cuda_v12 driver=12.9 pci_id=0000:01:00.0 type=discrete > total="2.0 GiB" available="1.8 GiB" > ``` yes, that's right, but the problem still exists
GiteaMirror commented 2025-11-12 14:46:01 -06:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 5, 2025):

And if you add a log that shows a model load, then the problem could be diagnosed.

@rick-github commented on GitHub (Nov 5, 2025): And if you add a log that shows a model load, then the problem could be diagnosed.
GiteaMirror commented 2025-11-12 14:46:02 -06:00
Author
Owner
Copy Link

@Savin1play140 commented on GitHub (Nov 5, 2025):

And if you add a log that shows a model load, then the problem could be diagnosed.

how i can do it?

@Savin1play140 commented on GitHub (Nov 5, 2025): > And if you add a log that shows a model load, then the problem could be diagnosed. how i can do it?
GiteaMirror commented 2025-11-12 14:46:02 -06:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 5, 2025):

  1. Load a model: ollama run qwen3:0.6b
  2. Add the log to this issue.
@rick-github commented on GitHub (Nov 5, 2025): 1. Load a model: `ollama run qwen3:0.6b` 2. Add the log to this issue.
GiteaMirror commented 2025-11-12 14:46:02 -06:00
Author
Owner
Copy Link

@Savin1play140 commented on GitHub (Nov 5, 2025):

  1. Load a model: ollama run qwen3:0.6b
  2. Add the log to this issue.

I already did this, judging by the sides, the model doesn't even try to load

@Savin1play140 commented on GitHub (Nov 5, 2025): > 1. Load a model: `ollama run qwen3:0.6b` > 2. Add the log to this issue. I already did this, judging by the sides, the model doesn't even try to load
GiteaMirror commented 2025-11-12 14:46:02 -06:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 5, 2025):

What was the response from ollama run qwen3:0.6b? That is, what was displayed on the terminal after running the command?

@rick-github commented on GitHub (Nov 5, 2025): What was the response from `ollama run qwen3:0.6b`? That is, what was displayed on the terminal after running the command?
GiteaMirror commented 2025-11-12 14:46:03 -06:00
Author
Owner
Copy Link

@Savin1play140 commented on GitHub (Nov 5, 2025):

What was the response from ollama run qwen3:0.6b? That is, what was displayed on the terminal after running the command?

eternal loading

@Savin1play140 commented on GitHub (Nov 5, 2025): > What was the response from `ollama run qwen3:0.6b`? That is, what was displayed on the terminal after running the command? eternal loading
GiteaMirror commented 2025-11-12 14:46:03 -06:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 5, 2025):

Could be #12699. Post the log.

@rick-github commented on GitHub (Nov 5, 2025): Could be #12699. Post the log.
GiteaMirror commented 2025-11-12 14:46:03 -06:00
Author
Owner
Copy Link

@Savin1play140 commented on GitHub (Nov 7, 2025):

The problem was solved in 0.12.11, thanks.

@Savin1play140 commented on GitHub (Nov 7, 2025): The problem was solved in 0.12.11, thanks.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
client2
cloud
compatibility
context-length
create
docker
documentation
embeddings
engine
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
install
integration
intel
js
linux
macos
memory
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
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-ollama#8567
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?