[GH-ISSUE #489] Can't build custom models #62262

Closed
opened 2026-05-03 08:01:26 -05:00 by GiteaMirror · 2 comments
Owner

Originally created by @selfagency on GitHub (Sep 7, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/489

Originally assigned to: @mxyng on GitHub.

I'm trying to build a custom model from a model I have already successfully pulled and used locally, but no matter what I do, it throws an error:

FROM codellama:latest
SYSTEM """

You are an expert in Nushell.
"""
❯ ollama list
NAME            	ID          	SIZE  	MODIFIED       
codellama:latest	36893bf9bc7f	3.8 GB	10 minutes ago	
llama2:13b      	984c614c4037	7.3 GB	24 minutes ago	
llama2:latest   	5c1a4ea68dd0	3.8 GB	7 minutes ago

❯ ollama create nushell -f ./Modelfile
parsing modelfile    
looking for model    
creating model system layer    
creating model license layer    
looking for model    
pulling model file    
⠋ pulling manifest  Error: pull model manifest: model not found

I'm running v0.0.18 installed via Homebrew on MacOS Sonoma Beta.

Originally created by @selfagency on GitHub (Sep 7, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/489 Originally assigned to: @mxyng on GitHub. I'm trying to build a custom model from a model I have already successfully pulled and used locally, but no matter what I do, it throws an error: ```Modelfile FROM codellama:latest SYSTEM """ You are an expert in Nushell. """ ``` ```sh ❯ ollama list NAME ID SIZE MODIFIED codellama:latest 36893bf9bc7f 3.8 GB 10 minutes ago llama2:13b 984c614c4037 7.3 GB 24 minutes ago llama2:latest 5c1a4ea68dd0 3.8 GB 7 minutes ago ❯ ollama create nushell -f ./Modelfile parsing modelfile looking for model creating model system layer creating model license layer looking for model pulling model file ⠋ pulling manifest Error: pull model manifest: model not found ``` I'm running v0.0.18 installed via Homebrew on MacOS Sonoma Beta.
Author
Owner

@mxyng commented on GitHub (Sep 7, 2023):

@selfagency, I'm not able to reproduce this on Homebrew's v0.0.18. Using your Modelfile, this should be the expected output:

$ cat Modelfile
FROM codellama:latest
SYSTEM """

You are an expert in Nushell.
"""
$ ollama create nushell -f ./Modelfile
parsing modelfile
looking for model
creating model system layer
creating config layer
using already created layer sha256:8268242df6f07932d5e6ef3ea276dd61a1420f0515a5f0ecdc7f11cbbf6ddb7c
using already created layer sha256:2c8743bdc4adb945dfa32717f767ce4984302fc802fc4fb80e235c67bc2df16d
using already created layer sha256:38fa20ee7daa67cfbdb181371e784bf30b22cd0c2bb12acc5b808a24b3e569f9
using already created layer sha256:578a2e81f7064c5118b93336dbe53dff6049bbeb4a8cee6c32a87579022e1aba
using already created layer sha256:404e21afdc6a34316cb23740018284bff1d16a480db6f8f59f4d5626dfb73376
writing layer sha256:9b21bb0bfe24e59a4bbf773ef7826b60ddcd2b478fd93d3213f205d5a62b95e8
writing layer sha256:7e969f73abb05a38023b880019a68fc69894aba8402552e1aad9ce8bf9568c76
writing manifest
success
$ brew info ollama
==> ollama: stable 0.0.18 (bottled), HEAD
Create, run, and share large language models (LLMs)
https://ollama.ai/
/opt/homebrew/Cellar/ollama/0.0.18 (7 files, 15.7MB) *
  Poured from bottle using the formulae.brew.sh API on 2023-09-07 at 15:57:57
From: https://github.com/Homebrew/homebrew-core/blob/HEAD/Formula/o/ollama.rb
License: MIT
==> Dependencies
Build: cmake ✘, go ✘
==> Options
--HEAD
        Install HEAD version
==> Caveats
To start ollama now and restart at login:
  brew services start ollama
Or, if you don't want/need a background service you can just run:
  /opt/homebrew/opt/ollama/bin/ollama serve
==> Analytics
install: 521 (30 days), 600 (90 days), 600 (365 days)
install-on-request: 521 (30 days), 600 (90 days), 600 (365 days)
build-error: 9 (30 days)

Can you attach the server logs? If the server is started with brew services, logs are located in $(brew --prefix)/var/log/ollama.log. If you started the server manually with ollama server, logs will be printed to stdout.

Is this install a fresh install or an upgrade? If it's an upgrade, it's possible it's running an old version of the server. If this is the case, restart the server and try again

<!-- gh-comment-id:1710873422 --> @mxyng commented on GitHub (Sep 7, 2023): @selfagency, I'm not able to reproduce this on Homebrew's v0.0.18. Using your Modelfile, this should be the expected output: ``` $ cat Modelfile FROM codellama:latest SYSTEM """ You are an expert in Nushell. """ $ ollama create nushell -f ./Modelfile parsing modelfile looking for model creating model system layer creating config layer using already created layer sha256:8268242df6f07932d5e6ef3ea276dd61a1420f0515a5f0ecdc7f11cbbf6ddb7c using already created layer sha256:2c8743bdc4adb945dfa32717f767ce4984302fc802fc4fb80e235c67bc2df16d using already created layer sha256:38fa20ee7daa67cfbdb181371e784bf30b22cd0c2bb12acc5b808a24b3e569f9 using already created layer sha256:578a2e81f7064c5118b93336dbe53dff6049bbeb4a8cee6c32a87579022e1aba using already created layer sha256:404e21afdc6a34316cb23740018284bff1d16a480db6f8f59f4d5626dfb73376 writing layer sha256:9b21bb0bfe24e59a4bbf773ef7826b60ddcd2b478fd93d3213f205d5a62b95e8 writing layer sha256:7e969f73abb05a38023b880019a68fc69894aba8402552e1aad9ce8bf9568c76 writing manifest success ``` ``` $ brew info ollama ==> ollama: stable 0.0.18 (bottled), HEAD Create, run, and share large language models (LLMs) https://ollama.ai/ /opt/homebrew/Cellar/ollama/0.0.18 (7 files, 15.7MB) * Poured from bottle using the formulae.brew.sh API on 2023-09-07 at 15:57:57 From: https://github.com/Homebrew/homebrew-core/blob/HEAD/Formula/o/ollama.rb License: MIT ==> Dependencies Build: cmake ✘, go ✘ ==> Options --HEAD Install HEAD version ==> Caveats To start ollama now and restart at login: brew services start ollama Or, if you don't want/need a background service you can just run: /opt/homebrew/opt/ollama/bin/ollama serve ==> Analytics install: 521 (30 days), 600 (90 days), 600 (365 days) install-on-request: 521 (30 days), 600 (90 days), 600 (365 days) build-error: 9 (30 days) ``` Can you attach the server logs? If the server is started with `brew services`, logs are located in `$(brew --prefix)/var/log/ollama.log`. If you started the server manually with `ollama server`, logs will be printed to stdout. Is this install a fresh install or an upgrade? If it's an upgrade, it's possible it's running an old version of the server. If this is the case, restart the server and try again
Author
Owner

@jmorganca commented on GitHub (Oct 23, 2023):

Hi there, thanks for creating an issue! Given there hasn't been much activity in the last month, I'll close this for now. Feel free to re-open it! Also, make sure to update to the latest version (0.1.3 as of this writing :-)

<!-- gh-comment-id:1775580662 --> @jmorganca commented on GitHub (Oct 23, 2023): Hi there, thanks for creating an issue! Given there hasn't been much activity in the last month, I'll close this for now. Feel free to re-open it! Also, make sure to update to the latest version (0.1.3 as of this writing :-)
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#62262