github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-16 05:05:37 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #1699] Modelfile parameters not set during creation #47471

New Issue
Closed
opened 2026-04-28 03:53:12 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-04-28 03:53:12 -05:00
Owner
Copy Link

Originally created by @tylertitsworth on GitHub (Dec 24, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1699

I have a model like so: (also providing system details)

$ cat /etc/os-release | head -n 4
PRETTY_NAME="Ubuntu 22.04.2 LTS"
NAME="Ubuntu"
VERSION_ID="22.04"
VERSION="22.04.2 LTS (Jammy Jellyfish)"
$ ollama -v
ollama version is 0.1.17
$ ollama show test --modelfile
# Modelfile generated by "ollama show"
# To build a new Modelfile based on this one, replace the FROM line with:
# FROM test:latest

FROM /usr/share/ollama/.ollama/models/blobs/sha256:8f7e99455a86ad490dab19d335d8c2d1d044e2744b923d4e78a4d84fe1457738
TEMPLATE """### System:
{{ .System }}

### User:
{{ .Prompt }}

### Assistant:
"""
SYSTEM """
Given the following extracted parts of a long document and a question, create a final answer with references ("SOURCES"). 
If you don't know the answer, just say that you don't know. Don't try to make up an answer.
ALWAYS return a "SOURCES" part in your answer.
---
Content: {context}
---
"""
PARAMETER repeat_penalty 1.3
PARAMETER temperature 0.4
PARAMETER top_k 20
PARAMETER top_p 0.65

My parameters shown in this output are the same as what I have set in my Modelfile.
However, when I query the API for the parameter values, I see that they are different values for repeat_penalty, temperature, and top_p:

$ ollama show test --parameters
repeat_penalty                 1
temperature                    0
top_k                          20
top_p                          1

$ curl http://localhost:11434/api/show -d '{"name": "test"}'
{
  "modelfile": "# Modelfile generated by \"ollama show\"\n# To build a new Modelfile based on this one, replace the FROM line with:\n# FROM test:latest\n\nFROM /usr/share/ollama/.ollama/models/blobs/sha256:8f7e99455a86ad490dab19d335d8c2d1d044e2744b923d4e78a4d84fe1457738\nTEMPLATE \"\"\"### System:\r\n{{ .System }}\r\n\r\n### User:\r\n{{ .Prompt }}\r\n\r\n### Assistant:\r\n\"\"\"\nSYSTEM \"\"\"\r\nGiven the following extracted parts of a long document and a question, create a final answer with references (\"SOURCES\"). \r\nIf you don't know the answer, just say that you don't know. Don't try to make up an answer.\r\nALWAYS return a \"SOURCES\" part in your answer.\r\n---\r\nContent: {context}\r\n---\r\n\"\"\"\nPARAMETER repeat_penalty 1.3\nPARAMETER temperature 0.4\nPARAMETER top_k 21\nPARAMETER top_p 0.65",
  "parameters": "top_p                          1\nrepeat_penalty                 1\ntemperature                    0\ntop_k                          20",
  "template": "### System:\r\n{{ .System }}\r\n\r\n### User:\r\n{{ .Prompt }}\r\n\r\n### Assistant:\r\n",
  "system": "\r\nGiven the following extracted parts of a long document and a question, create a final answer with references (\"SOURCES\"). \r\nIf you don't know the answer, just say that you don't know. Don't try to make up an answer.\r\nALWAYS return a \"SOURCES\" part in your answer.\r\n---\r\nContent: {context}\r\n---\r\n",
  "details":
    {
      "format": "gguf",
      "family": "llama",
      "families": ["llama"],
      "parameter_size": "7B",
      "quantization_level": "Q4_0",
    },
}

Cleaning up that JSON using Python:

>>> import requests
>>> defaults = dict(
...     pair.split()
...     for pair in requests.post(
...         "http://localhost:11434/api/show",
...         data='{"name": "test"}',
...         timeout=5
...     )
...     .json()["parameters"]
...     .split("\n")
...     if pair.strip()
... )
>>> defaults
{'top_p': '1', 'repeat_penalty': '1', 'temperature': '0', 'top_k': '20'}

I then changed my top_k value to make sure it just wasn't the same as the model default, and it did modify the value successfully.

If I change my input model from a GGUF model I quantized to neural-chat:latest, the problem is reproduced, just with some other default parameters:

$ ollama show test-neural-chat --parameters
repeat_penalty                 1
stop                           <|im_start|>
stop                           <|im_end|>
temperature                    0
top_k                          20
top_p                          1
num_ctx                        4096

For my sources, I'm following: https://github.com/jmorganca/ollama/blob/main/docs/modelfile.md#parameter

Originally created by @tylertitsworth on GitHub (Dec 24, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1699 I have a model like so: (also providing system details) ```Dockerfile $ cat /etc/os-release | head -n 4 PRETTY_NAME="Ubuntu 22.04.2 LTS" NAME="Ubuntu" VERSION_ID="22.04" VERSION="22.04.2 LTS (Jammy Jellyfish)" $ ollama -v ollama version is 0.1.17 $ ollama show test --modelfile # Modelfile generated by "ollama show" # To build a new Modelfile based on this one, replace the FROM line with: # FROM test:latest FROM /usr/share/ollama/.ollama/models/blobs/sha256:8f7e99455a86ad490dab19d335d8c2d1d044e2744b923d4e78a4d84fe1457738 TEMPLATE """### System: {{ .System }} ### User: {{ .Prompt }} ### Assistant: """ SYSTEM """ Given the following extracted parts of a long document and a question, create a final answer with references ("SOURCES"). If you don't know the answer, just say that you don't know. Don't try to make up an answer. ALWAYS return a "SOURCES" part in your answer. --- Content: {context} --- """ PARAMETER repeat_penalty 1.3 PARAMETER temperature 0.4 PARAMETER top_k 20 PARAMETER top_p 0.65 ``` My parameters shown in this output are the same as what I have set in my Modelfile. However, when I query the API for the parameter values, I see that they are different values for `repeat_penalty`, `temperature`, and `top_p`: ```bash $ ollama show test --parameters repeat_penalty 1 temperature 0 top_k 20 top_p 1 ``` ```json $ curl http://localhost:11434/api/show -d '{"name": "test"}' { "modelfile": "# Modelfile generated by \"ollama show\"\n# To build a new Modelfile based on this one, replace the FROM line with:\n# FROM test:latest\n\nFROM /usr/share/ollama/.ollama/models/blobs/sha256:8f7e99455a86ad490dab19d335d8c2d1d044e2744b923d4e78a4d84fe1457738\nTEMPLATE \"\"\"### System:\r\n{{ .System }}\r\n\r\n### User:\r\n{{ .Prompt }}\r\n\r\n### Assistant:\r\n\"\"\"\nSYSTEM \"\"\"\r\nGiven the following extracted parts of a long document and a question, create a final answer with references (\"SOURCES\"). \r\nIf you don't know the answer, just say that you don't know. Don't try to make up an answer.\r\nALWAYS return a \"SOURCES\" part in your answer.\r\n---\r\nContent: {context}\r\n---\r\n\"\"\"\nPARAMETER repeat_penalty 1.3\nPARAMETER temperature 0.4\nPARAMETER top_k 21\nPARAMETER top_p 0.65", "parameters": "top_p 1\nrepeat_penalty 1\ntemperature 0\ntop_k 20", "template": "### System:\r\n{{ .System }}\r\n\r\n### User:\r\n{{ .Prompt }}\r\n\r\n### Assistant:\r\n", "system": "\r\nGiven the following extracted parts of a long document and a question, create a final answer with references (\"SOURCES\"). \r\nIf you don't know the answer, just say that you don't know. Don't try to make up an answer.\r\nALWAYS return a \"SOURCES\" part in your answer.\r\n---\r\nContent: {context}\r\n---\r\n", "details": { "format": "gguf", "family": "llama", "families": ["llama"], "parameter_size": "7B", "quantization_level": "Q4_0", }, } ``` Cleaning up that JSON using Python: ```python >>> import requests >>> defaults = dict( ... pair.split() ... for pair in requests.post( ... "http://localhost:11434/api/show", ... data='{"name": "test"}', ... timeout=5 ... ) ... .json()["parameters"] ... .split("\n") ... if pair.strip() ... ) >>> defaults {'top_p': '1', 'repeat_penalty': '1', 'temperature': '0', 'top_k': '20'} ``` I then changed my top_k value to make sure it just wasn't the same as the model default, and it did modify the value successfully. If I change my input model from a GGUF model I quantized to `neural-chat:latest`, the problem is reproduced, just with some other default parameters: ```bash $ ollama show test-neural-chat --parameters repeat_penalty 1 stop <|im_start|> stop <|im_end|> temperature 0 top_k 20 top_p 1 num_ctx 4096 ``` For my sources, I'm following: https://github.com/jmorganca/ollama/blob/main/docs/modelfile.md#parameter
GiteaMirror added the bug label 2026-04-28 03:53:12 -05:00
GiteaMirror commented 2026-04-28 03:53:16 -05:00
Author
Owner
Copy Link

@pdevine commented on GitHub (Mar 12, 2024):

Hey @tylertitsworth sorry for the slow response. The API was just rounding the floats instead of properly displaying them. Thankfully this was actually fixed a while ago:

curl http://localhost:11434/api/show -d '{"name": "pdevine/testmodel"}' -s | jq ".parameters"
"repeat_penalty                 1.3\nstop                           \"[INST]\"\nstop                           \"[/INST]\"\nstop                           \"<<SYS>>\"\nstop                           \"<</SYS>>\""

Going to close out the issue.

<!-- gh-comment-id:1989690500 --> @pdevine commented on GitHub (Mar 12, 2024): Hey @tylertitsworth sorry for the slow response. The API was just rounding the floats instead of properly displaying them. Thankfully this was actually fixed a while ago: ``` curl http://localhost:11434/api/show -d '{"name": "pdevine/testmodel"}' -s | jq ".parameters" "repeat_penalty 1.3\nstop \"[INST]\"\nstop \"[/INST]\"\nstop \"<<SYS>>\"\nstop \"<</SYS>>\"" ``` Going to close out the issue.
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#47471
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?