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 #1103] Custom model repeats context in the response #551

New Issue
Closed
opened 2026-04-12 10:14:45 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-04-12 10:14:45 -05:00
Owner
Copy Link

Originally created by @sethmbhele on GitHub (Nov 12, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1103

Hello Friends

Firstly thank you so much for this amazing project. I have been playing around with it and having quite the blast learning the ins and outs of Ollama. If anyone can kindly assist with a challenge I am currently facing:
I created a Modelfile and passed temperature and system message; created and ran custom model. Everything works great and the new model is responding according to the new system message shared in the Modelfile. The challenge that I am facing is that on the second or third multi-turn chat, I am getting response + entire system message appended at the end. Any ideas on what I can try to fix this behaviour?
PS: Passing an instruction in the system message to NOT repeat system message in response did not work :(

Originally created by @sethmbhele on GitHub (Nov 12, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1103 Hello Friends Firstly thank you so much for this amazing project. I have been playing around with it and having quite the blast learning the ins and outs of Ollama. If anyone can kindly assist with a challenge I am currently facing: I created a Modelfile and passed temperature and system message; created and ran custom model. Everything works great and the new model is responding according to the new system message shared in the Modelfile. The challenge that I am facing is that on the second or third multi-turn chat, I am getting response + entire system message appended at the end. Any ideas on what I can try to fix this behaviour? PS: Passing an instruction in the system message to NOT repeat system message in response did not work :(
GiteaMirror commented 2026-04-12 10:14:47 -05:00
Author
Owner
Copy Link

@igorschlum commented on GitHub (Nov 12, 2023):

@sethmbhele

Hello, you are a step ahead of me. I would like to understand how to create and use a Modelfile. Could you share your experience with us?

<!-- gh-comment-id:1807258952 --> @igorschlum commented on GitHub (Nov 12, 2023): @sethmbhele Hello, you are a step ahead of me. I would like to understand how to create and use a Modelfile. Could you share your experience with us?
GiteaMirror commented 2026-04-12 10:14:47 -05:00
Author
Owner
Copy Link

@BruceMacD commented on GitHub (Nov 13, 2023):

This sounds like a job for stop tokens. They tell the LLM runner to stop generating when they're seen. Are you seeing <<SYS>> tags or anything like that in the output when it generates the system message again?

Here is an example of a modelfile with stop tokens:

FROM llama2
TEMPLATE """[INST] <<SYS>>{{ .System }}<</SYS>>

{{ .Prompt }} [/INST]
"""]
PARAMETER stop "[INST]"
PARAMETER stop "<<SYS>>"
<!-- gh-comment-id:1809165890 --> @BruceMacD commented on GitHub (Nov 13, 2023): This sounds like a job for stop tokens. They tell the LLM runner to stop generating when they're seen. Are you seeing `<<SYS>>` tags or anything like that in the output when it generates the system message again? Here is an example of a modelfile with stop tokens: ``` FROM llama2 TEMPLATE """[INST] <<SYS>>{{ .System }}<</SYS>> {{ .Prompt }} [/INST] """] PARAMETER stop "[INST]" PARAMETER stop "<<SYS>>" ```
GiteaMirror commented 2026-04-12 10:14:47 -05:00
Author
Owner
Copy Link

@sethmbhele commented on GitHub (Nov 19, 2023):

This sounds like a job for stop tokens. They tell the LLM runner to stop generating when they're seen. Are you seeing <<SYS>> tags or anything like that in the output when it generates the system message again?

Here is an example of a modelfile with stop tokens:

FROM llama2
TEMPLATE """[INST] <<SYS>>{{ .System }}<</SYS>>

{{ .Prompt }} [/INST]
"""]
PARAMETER stop "[INST]"
PARAMETER stop "<<SYS>>"

Thanks Bruce - this solved my issue

<!-- gh-comment-id:1817912457 --> @sethmbhele commented on GitHub (Nov 19, 2023): > This sounds like a job for stop tokens. They tell the LLM runner to stop generating when they're seen. Are you seeing `<<SYS>>` tags or anything like that in the output when it generates the system message again? > > Here is an example of a modelfile with stop tokens: > > ``` > FROM llama2 > TEMPLATE """[INST] <<SYS>>{{ .System }}<</SYS>> > > {{ .Prompt }} [/INST] > """] > PARAMETER stop "[INST]" > PARAMETER stop "<<SYS>>" > ``` Thanks Bruce - this solved my 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
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#551
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?