[GH-ISSUE #2200] ollama is stuck when translate language #27017

Closed
opened 2026-04-22 03:54:30 -05:00 by GiteaMirror · 2 comments
Owner

Originally created by @MunhakKim on GitHub (Jan 26, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/2200

ollama is stuck when i ask to translate language.
In my case, i put it in like '한글로' for prompt which means 'in Korean' after English output.
Sometimes ollama could translate perfectly and stable, but mostly ollama is stuck.
I think if there is a problem in text generation, It should be pause not is broken.
In this case, i have to restart ollama.

I guess it is about string format from streaming.

i am using docker image from docker hub.
using this command 'docker exec -it ollama ollama run mistral'
or api.

version: v0.1.21 (i have been tested from v0.1.18~)
model: mistral, mixtral latest, llama2 etc.

sorry for no logs.
But it is easy to cause problems when you put '한글로' in prompt.

Originally created by @MunhakKim on GitHub (Jan 26, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/2200 ollama is stuck when i ask to translate language. In my case, i put it in like '한글로' for prompt which means 'in Korean' after English output. Sometimes ollama could translate perfectly and stable, but mostly ollama is stuck. I think if there is a problem in text generation, It should be pause not is broken. In this case, **i have to restart ollama**. I guess it is about string format from streaming. i am using docker image from docker hub. using this command 'docker exec -it ollama ollama run mistral' or api. version: v0.1.21 (i have been tested from v0.1.18~) model: mistral, mixtral latest, llama2 etc. sorry for no logs. But it is easy to cause problems when you put '한글로' in prompt.
GiteaMirror added the question label 2026-04-22 03:54:30 -05:00
Author
Owner

@mxyng commented on GitHub (Mar 11, 2024):

Are you still having issues with this? I just tested that exact prompt with the latest release (0.1.29) on mistral without any hanging.

While the vanilla models are good with translation tasks, they're even better with a good system prompt. Also for reproducible results, I would suggest setting temperature to 0

<!-- gh-comment-id:1989193664 --> @mxyng commented on GitHub (Mar 11, 2024): Are you still having issues with this? I just tested that exact prompt with the latest release (0.1.29) on mistral without any hanging. While the vanilla models are good with translation tasks, they're even better with a good system prompt. Also for reproducible results, I would suggest setting temperature to 0
Author
Owner

@MunhakKim commented on GitHub (Mar 11, 2024):

No it was fixed this problem from 0.1.27(i guess). now i am happy to use ollama!! thank you for answer :)

<!-- gh-comment-id:1989618882 --> @MunhakKim commented on GitHub (Mar 11, 2024): No it was fixed this problem from 0.1.27(i guess). now i am happy to use ollama!! thank you for answer :)
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#27017