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 #15194] mlx runner failed #9726

New Issue
Open
opened 2026-04-12 22:36:35 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-04-12 22:36:35 -05:00
Owner
Copy Link

Originally created by @plainprince on GitHub (Apr 1, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/15194

What is the issue?

Downloaded the new ollama 0.19 and tried a prompt on qwen3.5:35b-a3b-coding-nvfp4 . After the AI did websearch it took about a minute before it gave the following error:
500 Internal Server Error: mlx runner failed: libc++abi: terminating due to uncaught exception of type std::runtime_error: [METAL] Command buffer execution failed: Impacting Interactivity (0000000e:kIOGPUCommandBufferCallbackErrorImpactingInteractivity)

Relevant log output

Searching for steganography techniques hide information in images text 2025…
Search results for steganography techniques hide information in images text 2025
Error: 500 Internal Server Error: mlx runner failed: libc++abi: terminating due to uncaught exception of type std::runtime_error: [METAL] Command buffer execution failed: Impacting Interactivity (0000000e:kIOGPUCommandBufferCallbackErrorImpactingInteractivity)

OS

macOS

GPU

Apple

CPU

Apple

Ollama version

0.19.0

Originally created by @plainprince on GitHub (Apr 1, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/15194 ### What is the issue? Downloaded the new ollama 0.19 and tried a prompt on qwen3.5:35b-a3b-coding-nvfp4 . After the AI did websearch it took about a minute before it gave the following error: 500 Internal Server Error: mlx runner failed: libc++abi: terminating due to uncaught exception of type std::runtime_error: [METAL] Command buffer execution failed: Impacting Interactivity (0000000e:kIOGPUCommandBufferCallbackErrorImpactingInteractivity) ### Relevant log output ```shell Searching for steganography techniques hide information in images text 2025… Search results for steganography techniques hide information in images text 2025 Error: 500 Internal Server Error: mlx runner failed: libc++abi: terminating due to uncaught exception of type std::runtime_error: [METAL] Command buffer execution failed: Impacting Interactivity (0000000e:kIOGPUCommandBufferCallbackErrorImpactingInteractivity) ``` ### OS macOS ### GPU Apple ### CPU Apple ### Ollama version 0.19.0
GiteaMirror added the bug label 2026-04-12 22:36:35 -05:00
GiteaMirror commented 2026-04-12 22:36:36 -05:00
Author
Owner
Copy Link

@iamkritika-official commented on GitHub (Apr 2, 2026):

Hey! I looked into this crash and found the root cause.

The macOS Metal GPU watchdog kills command buffers that take too long. With large models like qwen3.5:35b after a web search, the context gets very long and two things were happening:

  1. prefillChunkSize() was 2048 tokens — too large for a single Metal command buffer
  2. Generation loop was only syncing GPU every 256 tokens — letting buffers accumulate

I've raised a PR with a fix: reduced prefill chunk size to 512 and added periodic GPU sync every 64 tokens.

PR: https://github.com/iamkritika-official/ollama/pull/[YOUR_PR_NUMBER]

Can you test this on your machine and confirm if it resolves the crash?

<!-- gh-comment-id:4175022465 --> @iamkritika-official commented on GitHub (Apr 2, 2026): Hey! I looked into this crash and found the root cause. The macOS Metal GPU watchdog kills command buffers that take too long. With large models like qwen3.5:35b after a web search, the context gets very long and two things were happening: 1. prefillChunkSize() was 2048 tokens — too large for a single Metal command buffer 2. Generation loop was only syncing GPU every 256 tokens — letting buffers accumulate I've raised a PR with a fix: reduced prefill chunk size to 512 and added periodic GPU sync every 64 tokens. PR: https://github.com/iamkritika-official/ollama/pull/[YOUR_PR_NUMBER] Can you test this on your machine and confirm if it resolves the crash?
GiteaMirror commented 2026-04-12 22:36:37 -05:00
Author
Owner
Copy Link

@plainprince commented on GitHub (Apr 2, 2026):

https://github.com/iamkritika-official/ollama/pull/[YOUR_PR_NUMBER]

are you serious?
First of all, at least read your comment before you paste it from chatGPT
Second of all, what is the PR number?

<!-- gh-comment-id:4176163173 --> @plainprince commented on GitHub (Apr 2, 2026): > [https://github.com/iamkritika-official/ollama/pull/[YOUR_PR_NUMBER]](https://github.com/iamkritika-official/ollama/pull/%5BYOUR_PR_NUMBER%5D) are you serious? First of all, at least read your comment before you paste it from chatGPT Second of all, what is the PR number?
GiteaMirror commented 2026-04-12 22:36:37 -05:00
Author
Owner
Copy Link

@plainprince commented on GitHub (Apr 2, 2026):

Found the PR, #15206

<!-- gh-comment-id:4176168783 --> @plainprince commented on GitHub (Apr 2, 2026): Found the PR, #15206
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 bug
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#9726
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?