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 #665] Nonsense in output #26058

New Issue
Closed
opened 2026-04-22 01:56:27 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-22 01:56:27 -05:00
Owner
Copy Link

Originally created by @andersonbcdefg on GitHub (Oct 1, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/665

Originally assigned to: @BruceMacD on GitHub.

I've encountered this before when using llama.cpp on its own, but running ollama on my M2 I get random tokens sometimes in the output. (In particular, the model seems to like to produce the token "sierp". I can't be the only person who's experienced this. Not sure where this error comes from, but would be great if someone could figure it out and upstream it to GGML/Llama.cpp. Here's what it looks like:

Me: What's up?

Llama2: Hello! It's nice to meet you. sierp! How can I help you today? Do you
have any questions or topics you'd like to chat about?

Me: Repeat back your previous message.

Llama2: Sure, here is my previous message:

"Hello! It's nice to meet you. What's up?"

Would you like me to repeat anything else?

Originally created by @andersonbcdefg on GitHub (Oct 1, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/665 Originally assigned to: @BruceMacD on GitHub. I've encountered this before when using llama.cpp on its own, but running ollama on my M2 I get random tokens sometimes in the output. (In particular, the model seems to like to produce the token "sierp". I can't be the only person who's experienced this. Not sure where this error comes from, but would be great if someone could figure it out and upstream it to GGML/Llama.cpp. Here's what it looks like: Me: What's up? Llama2: Hello! It's nice to meet you. **sierp!** How can I help you today? Do you have any questions or topics you'd like to chat about? Me: Repeat back your previous message. Llama2: Sure, here is my previous message: "Hello! It's nice to meet you. What's up?" Would you like me to repeat anything else?
GiteaMirror added the bug label 2026-04-22 01:56:27 -05:00
GiteaMirror commented 2026-04-22 01:56:29 -05:00
Author
Owner
Copy Link

@willowell commented on GitHub (Oct 1, 2023):

I've seen this as well - sometimes Llama 2 7B Q4_0 mixes in non-English words in its first response after I say "Hello".

I don't recall seeing this behaviour when providing a system prompt like "Talk like a pirate."

Is it possible that the system prompt affects this behaviour somehow?

<!-- gh-comment-id:1742202063 --> @willowell commented on GitHub (Oct 1, 2023): I've seen this as well - sometimes Llama 2 7B Q4_0 mixes in non-English words in its first response after I say "Hello". I don't recall seeing this behaviour when providing a system prompt like "Talk like a pirate." Is it possible that the system prompt affects this behaviour somehow?
GiteaMirror commented 2026-04-22 01:56:30 -05:00
Author
Owner
Copy Link

@BruceMacD commented on GitHub (Oct 4, 2023):

This seems like it was a bug in the llama2 prompt template. I've updated it and I don't see the random output anymore.

You can get the updated prompt template by pulling:
ollama pull llama2

Let me know if you see any more issues.

<!-- gh-comment-id:1747601707 --> @BruceMacD commented on GitHub (Oct 4, 2023): This seems like it was a bug in the llama2 prompt template. I've updated it and I don't see the random output anymore. You can get the updated prompt template by pulling: `ollama pull llama2` Let me know if you see any more issues.
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#26058
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?