github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-08 08:59:13 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[PR #10185] [MERGED] Fix nondeterministic model unload order #23711

New Issue
Closed
opened 2026-04-19 17:09:51 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-04-19 17:09:51 -05:00
Owner
Copy Link

📋 Pull Request Information

Original PR: https://github.com/ollama/ollama/pull/10185
Author: @IreGaddr
Created: 4/9/2025
Status: ✅ Merged
Merged: 4/9/2025
Merged by: @dhiltgen

Base: main ← Head: fix-deterministic-unload

📝 Commits (1)

  • d5fe200 fix(scheduler): make model unload order deterministic

📊 Changes

1 file changed (+14 additions, -8 deletions)

View changed files

📝 server/sched.go (+14 -8)

📄 Description

Fix nondeterministic model unload order in scheduler
Summary
This pull request fixes a bug where Ollama's model scheduler sometimes failed to unload models predictably, leading to inconsistent memory release and occasional request failures, especially on Linux.

Root Cause
The scheduler iterated over Go maps (which are unordered) when selecting a model to unload. When multiple models had the same session duration, the unload choice was nondeterministic, varying across runs and platforms. This caused inconsistent unload behavior and resource management issues.

Fix
The unload selection logic was updated to:

Sort runners primarily by session duration (as before).
Add a secondary sort key: model path (string).
This ensures a deterministic unload order even when durations are equal, eliminating nondeterminism caused by map iteration order.

Impact
Model unload behavior is now consistent and predictable across platforms.
Fixes occasional failures to release models or accept new requests.
Tested on Linux with NVIDIA GPU, confirmed models unload deterministically and GPU memory is released as expected.

🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.

## 📋 Pull Request Information **Original PR:** https://github.com/ollama/ollama/pull/10185 **Author:** [@IreGaddr](https://github.com/IreGaddr) **Created:** 4/9/2025 **Status:** ✅ Merged **Merged:** 4/9/2025 **Merged by:** [@dhiltgen](https://github.com/dhiltgen) **Base:** `main` ← **Head:** `fix-deterministic-unload` --- ### 📝 Commits (1) - [`d5fe200`](https://github.com/ollama/ollama/commit/d5fe200a221a67130a8aec2db244def340dc863e) fix(scheduler): make model unload order deterministic ### 📊 Changes **1 file changed** (+14 additions, -8 deletions) <details> <summary>View changed files</summary> 📝 `server/sched.go` (+14 -8) </details> ### 📄 Description Fix nondeterministic model unload order in scheduler Summary This pull request fixes a bug where Ollama's model scheduler sometimes failed to unload models predictably, leading to inconsistent memory release and occasional request failures, especially on Linux. Root Cause The scheduler iterated over Go maps (which are unordered) when selecting a model to unload. When multiple models had the same session duration, the unload choice was nondeterministic, varying across runs and platforms. This caused inconsistent unload behavior and resource management issues. Fix The unload selection logic was updated to: Sort runners primarily by session duration (as before). Add a secondary sort key: model path (string). This ensures a deterministic unload order even when durations are equal, eliminating nondeterminism caused by map iteration order. Impact Model unload behavior is now consistent and predictable across platforms. Fixes occasional failures to release models or accept new requests. Tested on Linux with NVIDIA GPU, confirmed models unload deterministically and GPU memory is released as expected. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
GiteaMirror added the pull-request label 2026-04-19 17:09:51 -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 pull-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#23711
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?