github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 08:30:05 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #15266] Qwen 35b nvfp4 mlx Infinity thinking loop #71824

New Issue
Open
opened 2026-05-05 02:38:23 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-05-05 02:38:23 -05:00
Owner
Copy Link

Originally created by @Urcherd on GitHub (Apr 3, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/15266

What is the issue?

Command: ollama run qwen3.5:35b-a3b-coding-nvfp4 --verbose <<< "Write history in 500 lines"
Result: Error: mlx runner failed: time=2026-04-03T12:31:06.328+03:00 level=INFO source=cache.go:126 msg="cache hit" total=18 matched=18 cached=17 left=1
Env: Apple M4 Max | 64GB | Tahoe 26.3 (25D125)

Relevant log output

Server Log:
[GIN] 2026/04/03 - 12:31:06 | 200 |      53.625µs |       127.0.0.1 | HEAD     "/"
[GIN] 2026/04/03 - 12:31:06 | 200 |   85.233083ms |       127.0.0.1 | POST     "/api/show"
time=2026-04-03T12:31:06.327+03:00 level=INFO source=server.go:183 msg=ServeHTTP method=GET path=/v1/status took=14.833µs status="200 OK"
time=2026-04-03T12:31:06.328+03:00 level=INFO source=cache.go:126 msg="cache hit" total=18 matched=18 cached=17 left=1
[GIN] 2026/04/03 - 12:41:06 | 200 |         10m0s |       127.0.0.1 | POST     "/api/generate"
time=2026-04-03T12:41:06.335+03:00 level=INFO source=server.go:183 msg=ServeHTTP method=POST path=/v1/completions took=10m0.004317333s status="200 OK"
time=2026-04-03T12:41:06.419+03:00 level=INFO source=pipeline.go:55 msg="peak memory" size="21.20 GiB"
time=2026-04-03T12:41:06.419+03:00 level=INFO source=runner.go:149 msg="Request terminated" error="context canceled"

Output Log:
 *(Wait, I'll write the history now).*
    *(Okay).*
    *(Wait, I'll number them).*
    *(Okay).*
    *(Wait, I'll start).*
    *(Okay).*

    *(Wait, I need to be careful not to produce a wall of text that looks b
bad).*
    *(I'll make it readable).*
    *(Okay).*

    *(Wait, I'll start the output).*
    *(Okay).*
    *(Wait, I'll write the lines).*
    *(Okay).*
    *(Wait, I'll make sure I reach 500).*
    *(Okay).*

    *(Wait, I'll write the lines now).*
    *(Okay).*

    *(Wait, I'll start).*
    *(Okay).*

    *(Wait, I'll write the lines).*
    *(Okay).*

And last 3000 lines is:
    *(Wait, I'll write the lines).*
    *(Okay).*

OS

macOS

GPU

Apple

CPU

Apple

Ollama version

0.20.0

Originally created by @Urcherd on GitHub (Apr 3, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/15266 ### What is the issue? **Command:** ollama run qwen3.5:35b-a3b-coding-nvfp4 --verbose <<< "Write history in 500 lines" **Result:** Error: mlx runner failed: time=2026-04-03T12:31:06.328+03:00 level=INFO source=cache.go:126 msg="cache hit" total=18 matched=18 cached=17 left=1 Env: Apple M4 Max | 64GB | Tahoe 26.3 (25D125) ### Relevant log output ```shell Server Log: [GIN] 2026/04/03 - 12:31:06 | 200 | 53.625µs | 127.0.0.1 | HEAD "/" [GIN] 2026/04/03 - 12:31:06 | 200 | 85.233083ms | 127.0.0.1 | POST "/api/show" time=2026-04-03T12:31:06.327+03:00 level=INFO source=server.go:183 msg=ServeHTTP method=GET path=/v1/status took=14.833µs status="200 OK" time=2026-04-03T12:31:06.328+03:00 level=INFO source=cache.go:126 msg="cache hit" total=18 matched=18 cached=17 left=1 [GIN] 2026/04/03 - 12:41:06 | 200 | 10m0s | 127.0.0.1 | POST "/api/generate" time=2026-04-03T12:41:06.335+03:00 level=INFO source=server.go:183 msg=ServeHTTP method=POST path=/v1/completions took=10m0.004317333s status="200 OK" time=2026-04-03T12:41:06.419+03:00 level=INFO source=pipeline.go:55 msg="peak memory" size="21.20 GiB" time=2026-04-03T12:41:06.419+03:00 level=INFO source=runner.go:149 msg="Request terminated" error="context canceled" Output Log: *(Wait, I'll write the history now).* *(Okay).* *(Wait, I'll number them).* *(Okay).* *(Wait, I'll start).* *(Okay).* *(Wait, I need to be careful not to produce a wall of text that looks b bad).* *(I'll make it readable).* *(Okay).* *(Wait, I'll start the output).* *(Okay).* *(Wait, I'll write the lines).* *(Okay).* *(Wait, I'll make sure I reach 500).* *(Okay).* *(Wait, I'll write the lines now).* *(Okay).* *(Wait, I'll start).* *(Okay).* *(Wait, I'll write the lines).* *(Okay).* And last 3000 lines is: *(Wait, I'll write the lines).* *(Okay).* ``` ### OS macOS ### GPU Apple ### CPU Apple ### Ollama version 0.20.0
GiteaMirror added the bug label 2026-05-05 02:38:23 -05:00
GiteaMirror commented 2026-05-05 02:38:25 -05:00
Author
Owner
Copy Link

@andrenaP commented on GitHub (Apr 13, 2026):

Having the same issue using qwen3.5:latest

Output Log

system-promt="Keep your thinking process under 5 sentences."


Wait, I need to make sure I don't exceed 5 sentences in the thinking block.

Thinking block:

(doing thinking)

Okay, that's 5 sentences. I need to be under 5.

Let's combine.

Revised Thinking:

(doing thinking...)

That is 5 sentences. I need to be under 5 sentences (so max 4).

Let's merge 1 and 2.

(doing thinking...)

That's 3 sentences.

Wait, I need to output the code immediately after my initial plan.

The constraint says "Keep your thinking process under 5 sentences."

"Output the final code immediately after your initial plan."

So I write the thinking, then the code.

Okay, ready.

Wait, one more check: "Do not repeat logical checks."

Okay.

Wait, I need to make sure the code is valid.

(It keeps going until it loops always returning to `Wait, I need`)

OS

Arch Linux

GPU

Nvidia (4 GB VRAM)

CPU

Intel (16 GB RAM)

Ollama version

0.20.5

<!-- gh-comment-id:4234671521 --> @andrenaP commented on GitHub (Apr 13, 2026): Having the same issue using qwen3.5:latest ### Output Log `system-promt="Keep your thinking process under 5 sentences."` ```shell Wait, I need to make sure I don't exceed 5 sentences in the thinking block. Thinking block: (doing thinking) Okay, that's 5 sentences. I need to be under 5. Let's combine. Revised Thinking: (doing thinking...) That is 5 sentences. I need to be under 5 sentences (so max 4). Let's merge 1 and 2. (doing thinking...) That's 3 sentences. Wait, I need to output the code immediately after my initial plan. The constraint says "Keep your thinking process under 5 sentences." "Output the final code immediately after your initial plan." So I write the thinking, then the code. Okay, ready. Wait, one more check: "Do not repeat logical checks." Okay. Wait, I need to make sure the code is valid. (It keeps going until it loops always returning to `Wait, I need`) ``` ### OS Arch Linux ### GPU Nvidia (4 GB VRAM) ### CPU Intel (16 GB RAM) ### Ollama version 0.20.5
GiteaMirror commented 2026-05-05 02:38:25 -05:00
Author
Owner
Copy Link

@PureBlissAK commented on GitHub (Apr 18, 2026):

🤖 Automated Triage & Analysis Report

Issue: #15266
Analyzed: 2026-04-18T18:22:46.612020

Analysis

  • Type: unknown
  • Severity: medium
  • Components: unknown

Implementation Plan

  • Effort: medium
  • Steps:

This issue has been triaged and marked for implementation.

<!-- gh-comment-id:4274310601 --> @PureBlissAK commented on GitHub (Apr 18, 2026): <!-- ollama-issue-orchestrator:v1 issue:15266 --> ## 🤖 Automated Triage & Analysis Report **Issue**: #15266 **Analyzed**: 2026-04-18T18:22:46.612020 ### Analysis - **Type**: unknown - **Severity**: medium - **Components**: unknown ### Implementation Plan - **Effort**: medium - **Steps**: *This issue has been triaged and marked for implementation.*
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#71824
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?