github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-16 05:05:37 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #10345] The model's memory and GPU memory usage increases #84479

New Issue
Closed
opened 2026-05-09 20:51:16 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-05-09 20:51:16 -05:00
Owner
Copy Link

Originally created by @1556900941lizerui on GitHub (Apr 19, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/10345

I am running the qwq model on a server with L40GPU, but when I use ollama run qwq to run the model, I find that the model's memory usage needs 31GB, while on another server it only needs 23GB. When I adjust the context length to 8096, the model's memory usage reaches 68GB, causing 50% of the model parameters to be loaded into memory for CPU inference. On the other server, it only needs 32GB, which is the normal GPU memory usage. I'm not clear on the reason, it seems like I accidentally modified some parameters, causing the model to load into memory or GPU memory in a problematic way. Can you help me solve this confusion?

Originally created by @1556900941lizerui on GitHub (Apr 19, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/10345 I am running the qwq model on a server with L40GPU, but when I use ollama run qwq to run the model, I find that the model's memory usage needs 31GB, while on another server it only needs 23GB. When I adjust the context length to 8096, the model's memory usage reaches 68GB, causing 50% of the model parameters to be loaded into memory for CPU inference. On the other server, it only needs 32GB, which is the normal GPU memory usage. I'm not clear on the reason, it seems like I accidentally modified some parameters, causing the model to load into memory or GPU memory in a problematic way. Can you help me solve this confusion?
GiteaMirror commented 2026-05-09 20:51:17 -05:00
Author
Owner
Copy Link

@ALLMI78 commented on GitHub (Apr 19, 2025):

check

OLLAMA_NUM_PARALLEL
OLLAMA_MAX_LOADED_MODELS
OLLAMA_FLASH_ATTENTION

if it does not help you need to wait for the pros here ;)

<!-- gh-comment-id:2816760035 --> @ALLMI78 commented on GitHub (Apr 19, 2025): check OLLAMA_NUM_PARALLEL OLLAMA_MAX_LOADED_MODELS OLLAMA_FLASH_ATTENTION if it does not help you need to wait for the pros here ;)
GiteaMirror commented 2026-05-09 20:51:18 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Apr 19, 2025):

OLLAMA_NUM_PARALLEL

This is the likely cause. If unset, ollama uses a default of 1 or 4, depending on how much VRAM you have. I'm guessing that on the L40, ollama is using 4 and on the other server, 1.

<!-- gh-comment-id:2816766625 --> @rick-github commented on GitHub (Apr 19, 2025): > OLLAMA_NUM_PARALLEL This is the likely cause. If unset, ollama uses a default of 1 or 4, depending on how much VRAM you have. I'm guessing that on the L40, ollama is using 4 and on the other server, 1.
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#84479
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?