github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-08 08:59:13 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #13509] nemotron-3-nano loosing chat history #8905

New Issue
Closed
opened 2026-04-12 21:42:24 -05:00 by GiteaMirror · 5 comments
GiteaMirror commented 2026-04-12 21:42:24 -05:00
Owner
Copy Link

Originally created by @Digit-al on GitHub (Dec 17, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/13509

What is the issue?

When using nemotron, it sometime (often) loose the chat history.
I am using it through open-webui but the DEBUG in open-webui shows the history is passed to ollama, so I guess there is a problem with the parser.
See the example below.

chat_paris_weather.txt

Relevant log output

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.13.4

Originally created by @Digit-al on GitHub (Dec 17, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/13509 ### What is the issue? When using nemotron, it sometime (often) loose the chat history. I am using it through open-webui but the DEBUG in open-webui shows the history is passed to ollama, so I guess there is a problem with the parser. See the example below. [chat_paris_weather.txt](https://github.com/user-attachments/files/24211786/chat_paris_weather.txt) ### Relevant log output ```shell ``` ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.13.4
GiteaMirror added the bug label 2026-04-12 21:42:24 -05:00
GiteaMirror commented 2026-04-12 21:42:26 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Dec 17, 2025):

Increase the size of the context window.

<!-- gh-comment-id:3665144031 --> @rick-github commented on GitHub (Dec 17, 2025): Increase the size of the [context window](https://github.com/ollama/ollama/blob/main/docs/faq.mdx#how-can-i-specify-the-context-window-size).
GiteaMirror commented 2026-04-12 21:42:27 -05:00
Author
Owner
Copy Link

@Digit-al commented on GitHub (Dec 17, 2025):

That's odd, the num_ctx was 16384 so I thought it was enough for such a "small" exchange, wasn't it ?
I can try bigger context size but with Q4_K_M, it doesn't fit in my VRAM (1% on CPU with 16384 leading to approx. 57t/s and 5% on CPU with 32768 decreasing t/s to 23!).
32768 makes it, but will you consider proposing Q4_K_XL like the unsloth ones so all fits into 24GB VRAM ?

<!-- gh-comment-id:3665362038 --> @Digit-al commented on GitHub (Dec 17, 2025): That's odd, the num_ctx was 16384 so I thought it was enough for such a "small" exchange, wasn't it ? I can try bigger context size but with Q4_K_M, it doesn't fit in my VRAM (1% on CPU with 16384 leading to approx. 57t/s and 5% on CPU with 32768 decreasing t/s to 23!). 32768 makes it, but will you consider proposing Q4_K_XL like the unsloth ones so all fits into 24GB VRAM ?
GiteaMirror commented 2026-04-12 21:42:28 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Dec 17, 2025):

Just the response from the search_web tool call is ~21000 tokens.

If you want to use a Q4_K_XL quant:

ollama pull hf.co/unsloth/Nemotron-3-Nano-30B-A3B-GGUF:Q4_K_XL
echo FROM hf.co/unsloth/Nemotron-3-Nano-30B-A3B-GGUF:Q4_K_XL > Modelfile
ollama show --modelfile nemotron-3-nano:latest | grep -v "^FROM " >> Modelfile
ollama create nemotron-3-nano:30b-a3b-q4_K_XL
<!-- gh-comment-id:3667548532 --> @rick-github commented on GitHub (Dec 17, 2025): Just the response from the `search_web` tool call is ~21000 tokens. If you want to use a Q4_K_XL quant: ``` ollama pull hf.co/unsloth/Nemotron-3-Nano-30B-A3B-GGUF:Q4_K_XL echo FROM hf.co/unsloth/Nemotron-3-Nano-30B-A3B-GGUF:Q4_K_XL > Modelfile ollama show --modelfile nemotron-3-nano:latest | grep -v "^FROM " >> Modelfile ollama create nemotron-3-nano:30b-a3b-q4_K_XL ```
GiteaMirror commented 2026-04-12 21:42:31 -05:00
Author
Owner
Copy Link

@ParthSareen commented on GitHub (Dec 18, 2025):

@Digit-al I'd recommend doing some context engineering work for tools in order to only have the most useful (you pick what that means in your setting) information being passed in to the model. The most simple would be just to truncate the tool output.

<!-- gh-comment-id:3668726528 --> @ParthSareen commented on GitHub (Dec 18, 2025): @Digit-al I'd recommend doing some context engineering work for tools in order to only have the most useful (you pick what that means in your setting) information being passed in to the model. The most simple would be just to truncate the tool output.
GiteaMirror commented 2026-04-12 21:42:32 -05:00
Author
Owner
Copy Link

@Digit-al commented on GitHub (Dec 18, 2025):

Thanks a lot for your help and suggestions!

<!-- gh-comment-id:3672625751 --> @Digit-al commented on GitHub (Dec 18, 2025): Thanks a lot for your help and suggestions!
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#8905
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?