github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-21 13:11:39 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #549] Models sometimes prompt themselves #253

New Issue
Closed
opened 2026-04-12 09:46:42 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-04-12 09:46:42 -05:00
Owner
Copy Link

Originally created by @txstc55 on GitHub (Sep 18, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/549

I'm using uncensored model, the issue happened with uncensored-latest, uncensored 70b and any other uncensored model. Sometimes when I prompt the model, after it made a response, it will prompt itself with something like:

### Input:
something that generated by the model itself

### Response:
something that is a response to the input

This happens randomly and sometimes the ### Input tag becomes ### human tag
Any idea why this happens?

Originally created by @txstc55 on GitHub (Sep 18, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/549 I'm using uncensored model, the issue happened with uncensored-latest, uncensored 70b and any other uncensored model. Sometimes when I prompt the model, after it made a response, it will prompt itself with something like: ``` ### Input: something that generated by the model itself ### Response: something that is a response to the input ``` This happens randomly and sometimes the ### Input tag becomes ### human tag Any idea why this happens?
GiteaMirror commented 2026-04-12 09:46:44 -05:00
Author
Owner
Copy Link

@BruceMacD commented on GitHub (Sep 18, 2023):

It looks like there could be stop words missing from the default llama2-uncensored modelfile, these tell the LLM when to stop generating more text.

As a workaround until this gets fixed you can create your own llama2-uncensored Modelfile with the correct stop words. Here is how you do that:

  1. Create the Modelfile.
FROM llama2-uncensored:latest
TEMPLATE """### HUMAN:
{{ .Prompt }}

### RESPONSE:
"""
PARAMETER stop "### Input:"
PARAMETER stop "### Response:"
PARAMETER stop "### human"
  1. Load the custom model into Ollama via the CLI:
$ ollama create llama2-uncensored:custom -f path/to/Modelfile
  1. Now you can run it and the generation should stop when stop patterns are detected.
$ ollama run llama2-uncensored:custom
>>> hello
Hello back!
<!-- gh-comment-id:1723620641 --> @BruceMacD commented on GitHub (Sep 18, 2023): It looks like there could be `stop` words missing from the default llama2-uncensored modelfile, these tell the LLM when to stop generating more text. As a workaround until this gets fixed you can create your own llama2-uncensored Modelfile with the correct stop words. Here is how you do that: 1. Create the Modelfile. ``` FROM llama2-uncensored:latest TEMPLATE """### HUMAN: {{ .Prompt }} ### RESPONSE: """ PARAMETER stop "### Input:" PARAMETER stop "### Response:" PARAMETER stop "### human" ``` 2. Load the custom model into Ollama via the CLI: ``` $ ollama create llama2-uncensored:custom -f path/to/Modelfile ``` 3. Now you can run it and the generation should stop when stop patterns are detected. ``` $ ollama run llama2-uncensored:custom >>> hello Hello back! ```
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#253
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?