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 #6246] Modelfile - Customize a prompt #65944

New Issue
Closed
opened 2026-05-03 23:16:59 -05:00 by GiteaMirror · 11 comments
GiteaMirror commented 2026-05-03 23:16:59 -05:00
Owner
Copy Link

Originally created by @LucasFreitas88 on GitHub (Aug 7, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/6246

What is the issue?

I cannot change the modelfile file with new prompt instructions, as in the documentation example (Customize a prompt item).

Model: Llama 3.1 8B

Notebook: Macbook Air M1 - Mac Os Sonoma 14.6.1

The answer to the question posed in the example (hi) is an endless sequence of strange characters like:

H114@(02228'6.:@6@?/B+:&H((,/1A:8/=>=<.C-.C2>9C;H!$C=+5&,'&C7@44D>&BC=D"C%6<BB%;;$//2<D814),';?:@!!9:2H114@(02228'6.:@6@?/B+:&H((,/1A:8/=>=<.-C-.C2>9C;H!$C=+5&,'*&C7@44D>&BC=D"C%6< ...

What is happening? The procedure I used is exactly the same as described in the documentation.

OS

macOS

GPU

No response

CPU

No response

Ollama version

0.3.4

Originally created by @LucasFreitas88 on GitHub (Aug 7, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/6246 ### What is the issue? I cannot change the modelfile file with new prompt instructions, as in the documentation example (Customize a prompt item). Model: Llama 3.1 8B Notebook: Macbook Air M1 - Mac Os Sonoma 14.6.1 The answer to the question posed in the example (hi) is an endless sequence of strange characters like: H*114@(02228'6.:@6@?/B+:&H((,/1A:8/=>=<.C-.C2>9C*;H!$C=+5&,'*&C7@44D>&BC=D"C%6<BB%;;$*//2<D814),';?:@!!9:2H*114@(02228'6.:@6@?/B+:&H((,/1A:8/=>=<.-C-.C2>9C*;H!$C=+5&,'*&C7@44D>&BC=D"C%6< ... What is happening? The procedure I used is exactly the same as described in the documentation. ### OS macOS ### GPU _No response_ ### CPU _No response_ ### Ollama version 0.3.4
GiteaMirror added the bug label 2026-05-03 23:16:59 -05:00
GiteaMirror commented 2026-05-03 23:17:03 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Aug 8, 2024):

What's the contents of your modelfile?

<!-- gh-comment-id:2274588967 --> @rick-github commented on GitHub (Aug 8, 2024): What's the contents of your modelfile?
GiteaMirror commented 2026-05-03 23:17:03 -05:00
Author
Owner
Copy Link

@LucasFreitas88 commented on GitHub (Aug 8, 2024):

I tested it in two ways:

  1. just with the example content in the documentation

  2. with the same content as the original modelfile, plus a SYSTEM line with the documentation example prompt

<!-- gh-comment-id:2274714308 --> @LucasFreitas88 commented on GitHub (Aug 8, 2024): I tested it in two ways: 1) just with the example content in the documentation 2) with the same content as the original modelfile, plus a SYSTEM line with the documentation example prompt
GiteaMirror commented 2026-05-03 23:17:04 -05:00
Author
Owner
Copy Link

@LucasFreitas88 commented on GitHub (Aug 8, 2024):

both work with other models, such as mistral

<!-- gh-comment-id:2274716911 --> @LucasFreitas88 commented on GitHub (Aug 8, 2024): both work with other models, such as mistral
GiteaMirror commented 2026-05-03 23:17:04 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Aug 8, 2024):

There's another report (#6249) of llama3.1 on a Mac outputting random characters. If you run the model without a customized prompt, does it work any better? Have you tried other quants?

<!-- gh-comment-id:2275625121 --> @rick-github commented on GitHub (Aug 8, 2024): There's another report (#6249) of llama3.1 on a Mac outputting random characters. If you run the model without a customized prompt, does it work any better? Have you tried other quants?
GiteaMirror commented 2026-05-03 23:17:04 -05:00
Author
Owner
Copy Link

@LucasFreitas88 commented on GitHub (Aug 8, 2024):

I uploaded the version of ollama on my Mac yesterday, exclusively to test prompt changes in Llama3.1 modelfile.
Everything was fine until the day before yesterday, that's why I hadn't tested the model myself. I tested it now, on a new installation, and the model itself actually generates the same random characters.

<!-- gh-comment-id:2275909964 --> @LucasFreitas88 commented on GitHub (Aug 8, 2024): I uploaded the version of ollama on my Mac yesterday, exclusively to test prompt changes in Llama3.1 modelfile. Everything was fine until the day before yesterday, that's why I hadn't tested the model myself. I tested it now, on a new installation, and the model itself actually generates the same random characters.
GiteaMirror commented 2026-05-03 23:17:05 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Aug 8, 2024):

It's been reported that llama3.1:8b-instruct-q4_K_M works fine, do you find the same?

<!-- gh-comment-id:2275917455 --> @rick-github commented on GitHub (Aug 8, 2024): It's been reported that `llama3.1:8b-instruct-q4_K_M` works fine, do you find the same?
GiteaMirror commented 2026-05-03 23:17:06 -05:00
Author
Owner
Copy Link

@LucasFreitas88 commented on GitHub (Aug 8, 2024):

on my old installation (version before 0.3.4), the Llama3.1 was working perfectly without prompt changes.

<!-- gh-comment-id:2275918025 --> @LucasFreitas88 commented on GitHub (Aug 8, 2024): on my old installation (version before 0.3.4), the Llama3.1 was working perfectly without prompt changes.
GiteaMirror commented 2026-05-03 23:17:07 -05:00
Author
Owner
Copy Link

@LucasFreitas88 commented on GitHub (Aug 8, 2024):

llama3.1:8b-instruct-q4_K_M works fine

<!-- gh-comment-id:2275970341 --> @LucasFreitas88 commented on GitHub (Aug 8, 2024): llama3.1:8b-instruct-q4_K_M works fine
GiteaMirror commented 2026-05-03 23:17:08 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Aug 8, 2024):

Can you post server logs from an attempt to use llama3.1:8b? I'd like to see if you are getting the same error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory) messages as #6257.

<!-- gh-comment-id:2275992216 --> @rick-github commented on GitHub (Aug 8, 2024): Can you post server logs from an attempt to use `llama3.1:8b`? I'd like to see if you are getting the same `error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)` messages as #6257.
GiteaMirror commented 2026-05-03 23:17:08 -05:00
Author
Owner
Copy Link

@LucasFreitas88 commented on GitHub (Aug 8, 2024):

It was exactly this error! Thank you so much! I switched to a machine with 32GB of RAM, and it worked perfectly. I realized that it’s necessary to have 4-6GB free, at least on my Mac, to use Llama 3.1 8B, and there are only 8GB of RAM in total.

<!-- gh-comment-id:2276419342 --> @LucasFreitas88 commented on GitHub (Aug 8, 2024): It was exactly this error! Thank you so much! I switched to a machine with 32GB of RAM, and it worked perfectly. I realized that it’s necessary to have 4-6GB free, at least on my Mac, to use Llama 3.1 8B, and there are only 8GB of RAM in total.
GiteaMirror commented 2026-05-03 23:17:08 -05:00
Author
Owner
Copy Link

@LucasFreitas88 commented on GitHub (Aug 8, 2024):

Thanks!

<!-- gh-comment-id:2276419710 --> @LucasFreitas88 commented on GitHub (Aug 8, 2024): Thanks!
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#65944
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?