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 #10778] Ollama utilizes both the CPU and GPU simultaneously. #69138

New Issue
Closed
opened 2026-05-04 17:15:59 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-05-04 17:15:59 -05:00
Owner
Copy Link

Originally created by @qingningLime on GitHub (May 20, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/10778

When running Qwen2.5VL with Ollama using ROCm, Ollama utilizes both the CPU and GPU, whereas other models do not exhibit this issue.

When using Qwen2.5VL for image recognition with Ollama, both the CPU and GPU are utilized—even though GPU usage is visible, Ollama only returns the result after the CPU completes its tasks. This issue does not occur with other models (e.g., MiniCPM-V).

Image Image

Given that many modern vision-language models now support video understanding, but Ollama currently only handles image input, is video input functionality being considered for future updates? This would align with models like Qwen-2.5VL and others that are expanding into temporal visual data processing.

Originally created by @qingningLime on GitHub (May 20, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/10778 When running Qwen2.5VL with Ollama using ROCm, Ollama utilizes both the CPU and GPU, whereas other models do not exhibit this issue. When using Qwen2.5VL for image recognition with Ollama, both the CPU and GPU are utilized—even though GPU usage is visible, Ollama only returns the result after the CPU completes its tasks. This issue does not occur with other models (e.g., MiniCPM-V). <img width="862" alt="Image" src="https://github.com/user-attachments/assets/dacdeb7f-a5ab-41f1-923c-0f553ce31161" /> <img width="1277" alt="Image" src="https://github.com/user-attachments/assets/d185074f-ced7-4b24-8c89-84f977675743" /> Given that many modern vision-language models now support video understanding, but Ollama currently only handles image input, is video input functionality being considered for future updates? This would align with models like Qwen-2.5VL and others that are expanding into temporal visual data processing.
GiteaMirror commented 2026-05-04 17:16:05 -05:00
Author
Owner
Copy Link

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

qwen2.5vl requires more VRAM than minicpm-v, so part of the model is loaded in system RAM where the CPU does inference.

<!-- gh-comment-id:2893925851 --> @rick-github commented on GitHub (May 20, 2025): qwen2.5vl requires more VRAM than minicpm-v, so part of the model is loaded in system RAM where the CPU does inference.
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#69138
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?