[GH-ISSUE #2421] Offline models are not appearing on the Ollama server list #79091

Closed
opened 2026-05-09 03:57:06 -05:00 by GiteaMirror · 5 comments
Owner

Originally created by @HariKunapareddy on GitHub (Feb 9, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/2421

Problem
I download model on the the machine where Ollama installed and have internet access. Then moved the model files from the folder usr/share/ollama/.ollama/models to the new machine which doesn’t have internet access. I could see Ollama is not detecting those models and they are not visible as part of list command.

Expected

Ollama list

should list the new models files transferred.

Originally created by @HariKunapareddy on GitHub (Feb 9, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/2421 **Problem** I download model on the the machine where Ollama installed and have internet access. Then moved the model files from the folder usr/share/ollama/.ollama/models to the new machine which doesn’t have internet access. I could see Ollama is not detecting those models and they are not visible as part of list command. **Expected** Ollama list should list the new models files transferred.
Author
Owner

@jithinmukundan commented on GitHub (Feb 12, 2024):

Could you find any resolution? I am facing the same issue. Tried all possible changes related to permissions, systemd file etc. No luck yet.

<!-- gh-comment-id:1938941277 --> @jithinmukundan commented on GitHub (Feb 12, 2024): Could you find any resolution? I am facing the same issue. Tried all possible changes related to permissions, systemd file etc. No luck yet.
Author
Owner

@GanizaniSitara commented on GitHub (Feb 29, 2024):

What it initially succeeds with is "ollama cp my_invisble_model my_invisible_model2"

It creates the new folder and copies the manifest, but still doesn't list the model and when you try to run it insists on connecting to the internet. There must be somewhere else it looks?

So ...

ran a Window ProcMon figured out which file it touches in the BLOBS directory. Opened it up. It had bunch of SHA signatures which corresponded to files on the internet connected machine. One was Instruction template, the other one was license. Looking at timestamps, those two I've had on my internet connected machine for a while. I've popped them across and the model is now in

Ollama list

on the offline machine

<!-- gh-comment-id:1970180629 --> @GanizaniSitara commented on GitHub (Feb 29, 2024): What it initially succeeds with is "ollama cp my_invisble_model my_invisible_model2" It creates the new folder and copies the manifest, but still doesn't list the model and when you try to run it insists on connecting to the internet. There must be somewhere else it looks? So ... ran a Window ProcMon figured out which file it touches in the BLOBS directory. Opened it up. It had bunch of SHA signatures which corresponded to files on the internet connected machine. One was Instruction template, the other one was license. Looking at timestamps, those two I've had on my internet connected machine for a while. I've popped them across and the model is now in Ollama list on the offline machine
Author
Owner

@pdevine commented on GitHub (Mar 7, 2024):

Hey guys, sorry for the slow response.

If you are moving between Windows and Linux, the problem is the filenames of the blobs in linux look like sha256:<sha256 hash> whereas NTFS doesn't support a : in the file name. Unfortunately you have to rename the files.

For moving files to a different place, you should be able to use the OLLAMA_MODELS environment variable when you're starting Ollama. There is an FAQ which covers how to use that.

<!-- gh-comment-id:1982717459 --> @pdevine commented on GitHub (Mar 7, 2024): Hey guys, sorry for the slow response. If you are moving between Windows and Linux, the problem is the filenames of the blobs in linux look like `sha256:<sha256 hash>` whereas NTFS doesn't support a `:` in the file name. Unfortunately you have to rename the files. For moving files to a different place, you should be able to use the `OLLAMA_MODELS` environment variable when you're starting Ollama. There is an [FAQ](https://github.com/ollama/ollama/blob/main/docs/faq.md#where-are-models-stored) which covers how to use that.
Author
Owner

@GanizaniSitara commented on GitHub (Mar 7, 2024):

Hi Patrick, thank you, that's helpful for cross-platform moves. I've resolved mine from going windows to windows and making sure I copied all dependencies.

<!-- gh-comment-id:1984186451 --> @GanizaniSitara commented on GitHub (Mar 7, 2024): Hi Patrick, thank you, that's helpful for cross-platform moves. I've resolved mine from going windows to windows and making sure I copied all dependencies.
Author
Owner

@HariKunapareddy commented on GitHub (Mar 7, 2024):

Thanks! this solved the issue.

<!-- gh-comment-id:1984769760 --> @HariKunapareddy commented on GitHub (Mar 7, 2024): Thanks! this solved the issue.
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#79091