github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #14082] New context length base on vram size is crude rule(ollama0.15.5_rc3). Deepseek R1 32b can no longer works well(32Gvram ). #9194

New Issue
Closed
opened 2026-04-12 22:02:36 -05:00 by GiteaMirror · 6 comments
GiteaMirror commented 2026-04-12 22:02:36 -05:00
Owner
Copy Link

Originally created by @taozebra on GitHub (Feb 5, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/14082

What is the issue?

When i use deepseek r1 32b model.The output is very slow and the memory usage is abnormal under a 32G VRAM GPU.I think the new rule of vram use is inappropriate.

Relevant log output

OS

Linux

GPU

AMD

CPU

AMD

Ollama version

0.15.5rc3

Originally created by @taozebra on GitHub (Feb 5, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/14082 ### What is the issue? When i use deepseek r1 32b model.The output is very slow and the memory usage is abnormal under a 32G VRAM GPU.I think the new rule of vram use is inappropriate. ### Relevant log output ```shell ``` ### OS Linux ### GPU AMD ### CPU AMD ### Ollama version 0.15.5rc3
GiteaMirror added the bug label 2026-04-12 22:02:36 -05:00
GiteaMirror commented 2026-04-12 22:02:37 -05:00
Author
Owner
Copy Link

@taozebra commented on GitHub (Feb 5, 2026):

Image When i limit context lenth with 4096 in deepseekr1 output speed will be ok.
<!-- gh-comment-id:3850810814 --> @taozebra commented on GitHub (Feb 5, 2026): <img width="907" height="582" alt="Image" src="https://github.com/user-attachments/assets/82aba545-8db8-44f2-9fcb-80b625947a33" /> When i limit context lenth with 4096 in deepseekr1 output speed will be ok.
GiteaMirror commented 2026-04-12 22:02:37 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 5, 2026):

You can set OLLAMA_CONTEXT_LENGTH in the server environment, you don't need to modify the code. Or create a Modelfile that has PARAMETER num_ctx 4096.

<!-- gh-comment-id:3850875880 --> @rick-github commented on GitHub (Feb 5, 2026): You can set `OLLAMA_CONTEXT_LENGTH` in the server environment, you don't need to modify the code. Or create a Modelfile that has `PARAMETER num_ctx 4096`.
GiteaMirror commented 2026-04-12 22:02:38 -05:00
Author
Owner
Copy Link

@taozebra commented on GitHub (Feb 5, 2026):

You can set OLLAMA_CONTEXT_LENGTH in the server environment, you don't need to modify the code. Or create a Modelfile that has PARAMETER num_ctx 4096.

Most time I just use ollama by default setting. So when i update ollama to newest version i come across this issure. Thanks for your tip. they are usefully, In fact, I can modify openwebui configure of model to solve the problem(like Modefile). However, I still hope that ollama can be used out of the box.

<!-- gh-comment-id:3853178709 --> @taozebra commented on GitHub (Feb 5, 2026): > You can set `OLLAMA_CONTEXT_LENGTH` in the server environment, you don't need to modify the code. Or create a Modelfile that has `PARAMETER num_ctx 4096`. Most time I just use ollama by default setting. So when i update ollama to newest version i come across this issure. Thanks for your tip. they are usefully, In fact, I can modify openwebui configure of model to solve the problem(like Modefile). However, I still hope that ollama can be used out of the box.
GiteaMirror commented 2026-04-12 22:02:38 -05:00
Author
Owner
Copy Link

@jessegross commented on GitHub (Feb 5, 2026):

Can you please post your server logs? In theory, the VRAM usage should be pretty close to what you have.

<!-- gh-comment-id:3855271971 --> @jessegross commented on GitHub (Feb 5, 2026): Can you please post your server logs? In theory, the VRAM usage should be pretty close to what you have.
GiteaMirror commented 2026-04-12 22:02:38 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 5, 2026):

May be the num parallel issue.

<!-- gh-comment-id:3855638872 --> @rick-github commented on GitHub (Feb 5, 2026): May be the num parallel issue.
GiteaMirror commented 2026-04-12 22:02:39 -05:00
Author
Owner
Copy Link

@taozebra commented on GitHub (Feb 6, 2026):

May be the num parallel issue.

Yes, I set "OLLAMA_NUM_PARALLEL=2"

<!-- gh-comment-id:3857743561 --> @taozebra commented on GitHub (Feb 6, 2026): > May be the num parallel issue. Yes, I set "OLLAMA_NUM_PARALLEL=2"
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#9194
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?