github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 08:02:14 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #9726] Decouple Model Loading and Inference to Allow Dynamic Thread Configuration without Model Reload #6357

New Issue
Open
opened 2026-04-12 17:52:07 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-04-12 17:52:07 -05:00
Owner
Copy Link

Originally created by @luckycv on GitHub (Mar 13, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/9726

Description:

Currently, whenever we adjust configuration settings such as numThreads during inference, the system reloads the entire model. This can significantly slow down inference, especially when dealing with large models, since reloading the model is a time-consuming process.

Expected Behavior:

We would like to have the ability to modify inference-related configurations (such as numThreads) dynamically without the need to reload the model. This would allow us to adjust resource usage for inference on the fly without experiencing the overhead of reloading the model each time a configuration change is needed.

Proposed Solution:

Model Loading: Keep the model loading process independent of inference configurations. The model should be loaded once, and all configurations related to inference (e.g., thread count, batch size) should be changeable during the inference process.

Dynamic Configuration: Allow dynamic adjustment of inference parameters, such as numThreads, after the model is loaded, without requiring a reload of the model weights or the entire model state.

Use Case: This would benefit users running inference on large models, where adjusting resources (e.g., CPU threads) is often necessary for optimal performance. Having the ability to modify these settings without reloading the model would greatly improve both performance and user experience.

Additional Context:

Currently, the inference process and model loading are tightly coupled, making it difficult to adjust parameters like numThreads dynamically.
A solution to decouple these two processes would improve efficiency and reduce unnecessary overhead.

Originally created by @luckycv on GitHub (Mar 13, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/9726 **Description:** Currently, whenever we adjust configuration settings such as numThreads during inference, the system reloads the entire model. This can significantly slow down inference, especially when dealing with large models, since reloading the model is a time-consuming process. **Expected Behavior:** We would like to have the ability to modify inference-related configurations (such as numThreads) dynamically without the need to reload the model. This would allow us to adjust resource usage for inference on the fly without experiencing the overhead of reloading the model each time a configuration change is needed. **Proposed Solution:** Model Loading: Keep the model loading process independent of inference configurations. The model should be loaded once, and all configurations related to inference (e.g., thread count, batch size) should be changeable during the inference process. Dynamic Configuration: Allow dynamic adjustment of inference parameters, such as numThreads, after the model is loaded, without requiring a reload of the model weights or the entire model state. Use Case: This would benefit users running inference on large models, where adjusting resources (e.g., CPU threads) is often necessary for optimal performance. Having the ability to modify these settings without reloading the model would greatly improve both performance and user experience. **Additional Context:** Currently, the inference process and model loading are tightly coupled, making it difficult to adjust parameters like numThreads dynamically. A solution to decouple these two processes would improve efficiency and reduce unnecessary overhead.
GiteaMirror added the feature request label 2026-04-12 17:52:07 -05:00
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 feature request
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#6357
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?