github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-24 06:31:50 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[PR #10171] [MERGED] ollamarunner: Preallocate worst case graph at startup #59863

New Issue
Closed
opened 2026-04-29 14:47:15 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-04-29 14:47:15 -05:00
Owner
Copy Link

📋 Pull Request Information

Original PR: https://github.com/ollama/ollama/pull/10171
Author: @jessegross
Created: 4/8/2025
Status: ✅ Merged
Merged: 4/8/2025
Merged by: @jessegross

Base: main ← Head: jessegross/worst

📝 Commits (2)

  • 685c839 ggml: Check for OOM and return as Go errors
  • 0a3227b ollamarunner: Preallocate worst case graph at startup

📊 Changes

10 files changed (+181 additions, -55 deletions)

View changed files

📝 kvcache/cache.go (+3 -2)
📝 kvcache/causal.go (+39 -30)
📝 kvcache/causal_test.go (+5 -3)
📝 kvcache/encoder.go (+12 -3)
📝 kvcache/wrapper.go (+2 -2)
📝 ml/backend.go (+7 -0)
📝 ml/backend/ggml/ggml.go (+61 -13)
📝 model/model.go (+1 -1)
📝 runner/ollamarunner/cache_test.go (+1 -1)
📝 runner/ollamarunner/runner.go (+50 -0)

📄 Description

Currently, the KV cache and graph are lazily allocated as needed. The cache is fully allocated on first use of the corresponding layer whereas the graph grows with the size of the context.

This can be an issue if another application allocates more VRAM after we do our calculations - Ollama will crash in the middle of inference. If we instead allocate the maximum needed memory at startup of the runner, we will either succeed or fail at that point rather than at some surprising time in the future.

Currently, this only generates a worst case batch for text, which means that vision models may get a partial allocation and continue to lazily allocate the rest.

🔄 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/10171 **Author:** [@jessegross](https://github.com/jessegross) **Created:** 4/8/2025 **Status:** ✅ Merged **Merged:** 4/8/2025 **Merged by:** [@jessegross](https://github.com/jessegross) **Base:** `main` ← **Head:** `jessegross/worst` --- ### 📝 Commits (2) - [`685c839`](https://github.com/ollama/ollama/commit/685c839cda8b90144be3388dedec486247d563d8) ggml: Check for OOM and return as Go errors - [`0a3227b`](https://github.com/ollama/ollama/commit/0a3227bcebfa785d579a60b363894d2291a5829f) ollamarunner: Preallocate worst case graph at startup ### 📊 Changes **10 files changed** (+181 additions, -55 deletions) <details> <summary>View changed files</summary> 📝 `kvcache/cache.go` (+3 -2) 📝 `kvcache/causal.go` (+39 -30) 📝 `kvcache/causal_test.go` (+5 -3) 📝 `kvcache/encoder.go` (+12 -3) 📝 `kvcache/wrapper.go` (+2 -2) 📝 `ml/backend.go` (+7 -0) 📝 `ml/backend/ggml/ggml.go` (+61 -13) 📝 `model/model.go` (+1 -1) 📝 `runner/ollamarunner/cache_test.go` (+1 -1) 📝 `runner/ollamarunner/runner.go` (+50 -0) </details> ### 📄 Description Currently, the KV cache and graph are lazily allocated as needed. The cache is fully allocated on first use of the corresponding layer whereas the graph grows with the size of the context. This can be an issue if another application allocates more VRAM after we do our calculations - Ollama will crash in the middle of inference. If we instead allocate the maximum needed memory at startup of the runner, we will either succeed or fail at that point rather than at some surprising time in the future. Currently, this only generates a worst case batch for text, which means that vision models may get a partial allocation and continue to lazily allocate the rest. --- <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 2026-04-29 14:47:15 -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#59863
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?