github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-08 00:51:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #6797] Add the ability to remove a parameter using a Modelfile #50798

New Issue
Closed
opened 2026-04-28 17:07:57 -05:00 by GiteaMirror · 6 comments
GiteaMirror commented 2026-04-28 17:07:57 -05:00
Owner
Copy Link

Originally created by @dpkirchner on GitHub (Sep 13, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/6797

There's a model I'm interested in using with ollama that specifies a parameter no longer supported by ollama (or maybe llama.cpp). I'd like to be able to create a replacement with a Modelfile that overrides the parameter by removing it entirely, if possible.

(The specific model is codellama:7b-instruct and the parameter is rope_frequency_base, for what it's worth, but this might be generally useful.)

Originally created by @dpkirchner on GitHub (Sep 13, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/6797 There's a model I'm interested in using with ollama that specifies a parameter no longer supported by ollama (or maybe llama.cpp). I'd like to be able to create a replacement with a Modelfile that overrides the parameter by removing it entirely, if possible. (The specific model is `codellama:7b-instruct` and the parameter is `rope_frequency_base`, for what it's worth, but this might be generally useful.)
GiteaMirror added the feature request label 2026-04-28 17:07:57 -05:00
GiteaMirror commented 2026-04-28 17:07:59 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 13, 2024):

Can you not just delete the line in the modelfile?

$ ollama show --modelfile codellama:7b-instruct | grep -v rope_frequency_base | sed -e 's/^FROM.*/FROM codellama:7b-instruct/' > Modelfile.norope
$ ollama create codellama:7b-instruct-no-rope -f Modelfile.norope
$ ollama run codellama:7b-instruct-no-rope "write a python script to print numbers 1..10.  don't explain"
```
for i in range(1, 11):
    print(i)
```
<!-- gh-comment-id:2350715139 --> @rick-github commented on GitHub (Sep 13, 2024): Can you not just delete the line in the modelfile? ````console $ ollama show --modelfile codellama:7b-instruct | grep -v rope_frequency_base | sed -e 's/^FROM.*/FROM codellama:7b-instruct/' > Modelfile.norope $ ollama create codellama:7b-instruct-no-rope -f Modelfile.norope $ ollama run codellama:7b-instruct-no-rope "write a python script to print numbers 1..10. don't explain" ``` for i in range(1, 11): print(i) ``` ````
GiteaMirror commented 2026-04-28 17:08:00 -05:00
Author
Owner
Copy Link

@dpkirchner commented on GitHub (Sep 14, 2024):

Can you not just delete the line in the modelfile?

$ ollama show --modelfile codellama:7b-instruct | grep -v rope_frequency_base | sed -e 's/^FROM.*/FROM codellama:7b-instruct/' > Modelfile.norope
$ ollama create codellama:7b-instruct-no-rope -f Modelfile.norope
$ ollama run codellama:7b-instruct-no-rope "write a python script to print numbers 1..10.  don't explain"

First thing I tried -- when I do this, the new model still has that rope_frequency_base parameter:

root@desktop:/app/backend# ollama -v
ollama version is 0.3.10
root@desktop:/app/backend# head -n 20 Modelfile.norope 
# Modelfile generated by "ollama show"
# To build a new Modelfile based on this, replace FROM with:
# FROM codellama:7b-instruct

FROM codellama:7b-instruct
TEMPLATE "[INST] <<SYS>>{{ .System }}<</SYS>>

{{ .Prompt }} [/INST]
"
PARAMETER stop [INST]
PARAMETER stop [/INST]
PARAMETER stop <<SYS>>
PARAMETER stop <</SYS>>
LICENSE """LLAMA 2 COMMUNITY LICENSE AGREEMENT	
Llama 2 Version Release Date: July 18, 2023

"Agreement" means the terms and conditions for use, reproduction, distribution and 
modification of the Llama Materials set forth herein.

"Documentation" means the specifications, manuals and documentation 
root@desktop:/app/backend# ollama create codellama:7b-instruct-no-rope -f Modelfile.norope
transferring model data 
using existing layer sha256:3a43f93b78ec50f7c4e4dc8bd1cb3fff5a900e7d574c51a6f7495e48486e0dac 
using existing layer sha256:8c17c2ebb0ea011be9981cc3922db8ca8fa61e828c5d3f44cb6ae342bf80460b 
using existing layer sha256:590d74a5569b8a20eb2a8b0aa869d1d1d3faf6a7fdda1955ae827073c7f502fc 
using existing layer sha256:2e0493f67d0c8c9c68a8aeacdf6a38a2151cb3c4c1d42accf296e19810527988 
using existing layer sha256:8c17c2ebb0ea011be9981cc3922db8ca8fa61e828c5d3f44cb6ae342bf80460b 
using existing layer sha256:590d74a5569b8a20eb2a8b0aa869d1d1d3faf6a7fdda1955ae827073c7f502fc 
creating new layer sha256:cfac7756790d0eccff9c43784590bf23ecadb0466c8093cbb28aae91015ca3f4 
creating new layer sha256:94ab2b590e2fde14c3b139c2d4226be28b4d494dd82e48936bfff01cee4ea48e 
writing manifest 
success 
root@desktop:/app/backend# ollama show --modelfile codellama:7b-instruct-no-rope | grep rope_freq
PARAMETER rope_frequency_base 1e+06
<!-- gh-comment-id:2350725757 --> @dpkirchner commented on GitHub (Sep 14, 2024): > Can you not just delete the line in the modelfile? > > ``` > $ ollama show --modelfile codellama:7b-instruct | grep -v rope_frequency_base | sed -e 's/^FROM.*/FROM codellama:7b-instruct/' > Modelfile.norope > $ ollama create codellama:7b-instruct-no-rope -f Modelfile.norope > $ ollama run codellama:7b-instruct-no-rope "write a python script to print numbers 1..10. don't explain" First thing I tried -- when I do this, the new model still has that `rope_frequency_base` parameter: ``` root@desktop:/app/backend# ollama -v ollama version is 0.3.10 root@desktop:/app/backend# head -n 20 Modelfile.norope # Modelfile generated by "ollama show" # To build a new Modelfile based on this, replace FROM with: # FROM codellama:7b-instruct FROM codellama:7b-instruct TEMPLATE "[INST] <<SYS>>{{ .System }}<</SYS>> {{ .Prompt }} [/INST] " PARAMETER stop [INST] PARAMETER stop [/INST] PARAMETER stop <<SYS>> PARAMETER stop <</SYS>> LICENSE """LLAMA 2 COMMUNITY LICENSE AGREEMENT Llama 2 Version Release Date: July 18, 2023 "Agreement" means the terms and conditions for use, reproduction, distribution and modification of the Llama Materials set forth herein. "Documentation" means the specifications, manuals and documentation root@desktop:/app/backend# ollama create codellama:7b-instruct-no-rope -f Modelfile.norope transferring model data using existing layer sha256:3a43f93b78ec50f7c4e4dc8bd1cb3fff5a900e7d574c51a6f7495e48486e0dac using existing layer sha256:8c17c2ebb0ea011be9981cc3922db8ca8fa61e828c5d3f44cb6ae342bf80460b using existing layer sha256:590d74a5569b8a20eb2a8b0aa869d1d1d3faf6a7fdda1955ae827073c7f502fc using existing layer sha256:2e0493f67d0c8c9c68a8aeacdf6a38a2151cb3c4c1d42accf296e19810527988 using existing layer sha256:8c17c2ebb0ea011be9981cc3922db8ca8fa61e828c5d3f44cb6ae342bf80460b using existing layer sha256:590d74a5569b8a20eb2a8b0aa869d1d1d3faf6a7fdda1955ae827073c7f502fc creating new layer sha256:cfac7756790d0eccff9c43784590bf23ecadb0466c8093cbb28aae91015ca3f4 creating new layer sha256:94ab2b590e2fde14c3b139c2d4226be28b4d494dd82e48936bfff01cee4ea48e writing manifest success root@desktop:/app/backend# ollama show --modelfile codellama:7b-instruct-no-rope | grep rope_freq PARAMETER rope_frequency_base 1e+06 ```
GiteaMirror commented 2026-04-28 17:08:00 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 14, 2024):

Ah, I see, the parameter is set in the k/v space of the model. The model appears to work with it set, and ollama ignores it. What's the use case for removing it?

<!-- gh-comment-id:2350739537 --> @rick-github commented on GitHub (Sep 14, 2024): Ah, I see, the parameter is set in the k/v space of the model. The model appears to work with it set, and ollama ignores it. What's the use case for removing it?
GiteaMirror commented 2026-04-28 17:08:01 -05:00
Author
Owner
Copy Link

@dpkirchner commented on GitHub (Sep 14, 2024):

It might just be an unexpected warning, however I'm not sure if it's affecting the output of requests. I'm using a vscode extension that sends the request (from debug logs):

time=2024-09-14T00:32:41.496Z level=DEBUG source=routes.go:1363 msg="chat request" images=0 prompt="[INST] <<SYS>>You are a helpful, respectful and honest coding assistant.\nAlways reply using markdown.\nBe clear and concise, prioritizing brevity in your responses.\nFor code refactoring, use markdown with appropriate code formatting.<</SYS>>\n\nWrite a python hello world [/INST]\n"

and the response is Sorry, I don't understand. Please try again.. I was looking for errors and saw the line about the rope option:

time=2024-09-14T00:32:41.805Z level=WARN source=types.go:509 msg="invalid option provided" option=rope_frequency_base

and thought that's why the request is failing. I'm not sure whether or not it is the cause.

If the invalid option is not the cause then I think this request can be closed.

<!-- gh-comment-id:2350743216 --> @dpkirchner commented on GitHub (Sep 14, 2024): It might just be an unexpected warning, however I'm not sure if it's affecting the output of requests. I'm using a vscode extension that sends the request (from debug logs): ``` time=2024-09-14T00:32:41.496Z level=DEBUG source=routes.go:1363 msg="chat request" images=0 prompt="[INST] <<SYS>>You are a helpful, respectful and honest coding assistant.\nAlways reply using markdown.\nBe clear and concise, prioritizing brevity in your responses.\nFor code refactoring, use markdown with appropriate code formatting.<</SYS>>\n\nWrite a python hello world [/INST]\n" ``` and the response is `Sorry, I don't understand. Please try again.`. I was looking for errors and saw the line about the rope option: ``` time=2024-09-14T00:32:41.805Z level=WARN source=types.go:509 msg="invalid option provided" option=rope_frequency_base ``` and thought that's why the request is failing. I'm not sure whether or not it is the cause. If the invalid option is not the cause then I think this request can be closed.
GiteaMirror commented 2026-04-28 17:08:01 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 14, 2024):

The invalid option should have no effect on the generation.

I ran the prompt you had several times and didn't see a failure:

$ curl -s localhost:11434/api/chat -d '{"model":"codellama:7b-instruct","messages":[{"role":"system","content":"You are a helpful, respectful and honest coding assistant.\nAlways reply using markdown.\nBe clear and concise, prioritizing brevity in your responses.\nFor code refactoring, use markdown with appropriate code formatting."},{"role":"user","content":"Write a python hello world"}],"stream":false}' | jq
{
  "model": "codellama:7b-instruct",
  "created_at": "2024-09-14T01:54:11.218041687Z",
  "message": {
    "role": "assistant",
    "content": "```python\nprint(\"Hello, World!\")\n```"
  },
  "done_reason": "stop",
  "done": true,
  "total_duration": 289237666,
  "load_duration": 6886725,
  "prompt_eval_count": 74,
  "prompt_eval_duration": 18785000,
  "eval_count": 13,
  "eval_duration": 127262000
}

Is it still failing for you? If not, I think that the failure is just one of those hazards of LLMs, like hallucinations - sometimes they just go off the rails.

<!-- gh-comment-id:2350780313 --> @rick-github commented on GitHub (Sep 14, 2024): The invalid option should have no effect on the generation. I ran the prompt you had several times and didn't see a failure: ```console $ curl -s localhost:11434/api/chat -d '{"model":"codellama:7b-instruct","messages":[{"role":"system","content":"You are a helpful, respectful and honest coding assistant.\nAlways reply using markdown.\nBe clear and concise, prioritizing brevity in your responses.\nFor code refactoring, use markdown with appropriate code formatting."},{"role":"user","content":"Write a python hello world"}],"stream":false}' | jq { "model": "codellama:7b-instruct", "created_at": "2024-09-14T01:54:11.218041687Z", "message": { "role": "assistant", "content": "```python\nprint(\"Hello, World!\")\n```" }, "done_reason": "stop", "done": true, "total_duration": 289237666, "load_duration": 6886725, "prompt_eval_count": 74, "prompt_eval_duration": 18785000, "eval_count": 13, "eval_duration": 127262000 } ``` Is it still failing for you? If not, I think that the failure is just one of those hazards of LLMs, like hallucinations - sometimes they just go off the rails.
GiteaMirror commented 2026-04-28 17:08:01 -05:00
Author
Owner
Copy Link

@dpkirchner commented on GitHub (Sep 14, 2024):

That curl works. Maybe it's just a problem with the extension after all. I latched on to the warning prematurely. Thanks for your help!

<!-- gh-comment-id:2351009028 --> @dpkirchner commented on GitHub (Sep 14, 2024): That curl works. Maybe it's just a problem with the extension after all. I latched on to the warning prematurely. Thanks for your help!
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 feature request
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#50798
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?