github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #8493] Long context for Qwen2.5 is possible but needs something to work #67527

New Issue
Open
opened 2026-05-04 10:40:06 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-05-04 10:40:06 -05:00
Owner
Copy Link

Originally created by @devlux76 on GitHub (Jan 20, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/8493

The instructions for Qwen2.5 (all of them) state quite clearly that everything from 7B on up have 128k context. However in order to use that context you need to do something...
https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct#processing-long-texts

For supported frameworks, you could add the following to config.json to enable YaRN:

{
  ...,
  "rope_scaling": {
    "factor": 4.0,
    "original_max_position_embeddings": 32768,
    "type": "yarn"
  }
}

What is the ollama method of achieving this because I'd really like to use these models at their full context length.
Thanks!

Originally created by @devlux76 on GitHub (Jan 20, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/8493 The instructions for Qwen2.5 (all of them) state quite clearly that everything from 7B on up have 128k context. However in order to use that context you need to do something... https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct#processing-long-texts For supported frameworks, you could add the following to config.json to enable YaRN: ``` { ..., "rope_scaling": { "factor": 4.0, "original_max_position_embeddings": 32768, "type": "yarn" } } ``` What is the ollama method of achieving this because I'd really like to use these models at their full context length. Thanks!
GiteaMirror commented 2026-05-04 10:40:09 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jan 20, 2025):

https://github.com/ggerganov/llama.cpp/pull/10698 for adding yarn support for qwen. The model needs to be re-quantized with llama.cpp.

We advise adding the rope_scaling configuration only when processing long contexts is required.

It seems it might degrade the performance for shorter context lengths so probably not appropriate for the default model.

<!-- gh-comment-id:2601891513 --> @rick-github commented on GitHub (Jan 20, 2025): https://github.com/ggerganov/llama.cpp/pull/10698 for adding yarn support for qwen. The model needs to be re-quantized with llama.cpp. > We advise adding the rope_scaling configuration only when processing long contexts is required. It seems it might degrade the performance for shorter context lengths so probably not appropriate for the default model.
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#67527
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?