github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-16 05:05:37 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #14170] Qwen2.5-VL 3B crashes with HTTP 500 on specific PNG image #55750

New Issue
Open
opened 2026-04-29 09:41:34 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-04-29 09:41:34 -05:00
Owner
Copy Link

Originally created by @megengels on GitHub (Feb 9, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/14170

What is the issue?

Bug: Qwen2.5-VL 3B crashes with HTTP 500 on specific PNG image

Environment

  • Ollama: v0.15.6 (latest)
  • OS: Ubuntu 24 (WSL2)
  • GPU: NVIDIA GeForce RTX 5080 (16GB VRAM)
  • Model: qwen2.5vl:3b

Description

Qwen2.5-VL 3B consistently crashes with a 500 error on a specific PNG image while a different image with the exact same dimensions works fine. The runner enters a crash-loop after the failure.

Reproduction

  1. Download the two test images attached below (synthetic payslips, fake data)
  2. Run the following for each image:
import base64, json, urllib.request

with open('IMAGE_PATH', 'rb') as f:
    img_b64 = base64.b64encode(f.read()).decode()

data = json.dumps({
    'model': 'qwen2.5vl:3b',
    'messages': [{'role': 'user', 'content': 'Quel est le net à payer ?', 'images': [img_b64]}],
    'stream': False
}).encode()

req = urllib.request.Request('http://localhost:11434/api/chat', data=data, headers={'Content-Type': 'application/json'})
resp = urllib.request.urlopen(req, timeout=60)
print(resp.read().decode())

Results

Image Size Dimensions Result
bulletin_LÉVY (modern layout) 7.15 MB 1654x2339 ✅ Works
bulletin_CHARRIER (classic layout) 4.62 MB 1654x2339 ❌ 500 crash

Observed behavior

  • After ~8.5 seconds, the runner crashes
  • Ollama enters a crash-loop, restarting the runner ~20 times:
[GIN] 500 | 8.496894644s | POST "/api/chat"
starting runner cmd="/usr/local/bin/ollama runner --ollama-engine --port 34119"
starting runner cmd="/usr/local/bin/ollama runner --ollama-engine --port 35553"
starting runner cmd="/usr/local/bin/ollama runner --ollama-engine --port 45419"
... (~20 more restarts)

Additional context

  • Both images are the same dimensions (1654x2339), crashing one is actually smaller (4.62 MB vs 7.15 MB)
  • Plenty of VRAM available (12+ GB free / 16 GB total)
  • No other processes running on GPU
  • Reproduced consistently across multiple attempts and service restarts
  • Test images are synthetic French payslips (fake data), safe to attach
Image Image

Relevant log output

Feb 09 14:18:57 MEGDIG-D-MES ollama[605557]: time=2026-02-09T14:18:57.793+01:00 level=INFO source=runner.go:464 msg="failure during GPU discovery" OLLAMA_LIBRARY_PATH="[/usr/local/lib/ollama /usr/local/lib/ollama/cuda_v13]" extra_envs=map[] error="failed to finish discovery before timeout"
Feb 09 14:18:57 MEGDIG-D-MES ollama[605557]: time=2026-02-09T14:18:57.793+01:00 level=WARN source=runner.go:356 msg="unable to refresh free memory, using old values"

Feb 09 14:23:03 MEGDIG-D-MES ollama[654807]: [GIN] 2026/02/09 - 14:23:03 | 500 |  8.496894644s |       127.0.0.1 | POST     "/api/chat"
Feb 09 14:24:06 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:06.604+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 34119"
Feb 09 14:24:07 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:07.069+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 35553"
Feb 09 14:24:07 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:07.314+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 45419"
Feb 09 14:24:07 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:07.564+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 44979"
Feb 09 14:24:07 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:07.814+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 37187"
Feb 09 14:24:08 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:08.064+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 45377"
Feb 09 14:24:08 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:08.314+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 33263"
Feb 09 14:24:08 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:08.564+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 38835"
Feb 09 14:24:08 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:08.814+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 42785"
Feb 09 14:24:09 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:09.064+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 38421"
Feb 09 14:24:09 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:09.314+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 36565"
Feb 09 14:24:09 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:09.564+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 34367"
Feb 09 14:24:09 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:09.814+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 40257"
Feb 09 14:24:10 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:10.064+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 44767"
Feb 09 14:24:10 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:10.314+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 45329"
Feb 09 14:24:10 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:10.564+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 34121"
Feb 09 14:24:10 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:10.814+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 36567"
Feb 09 14:24:11 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:11.065+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 35639"
Feb 09 14:24:11 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:11.314+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 36387"
Feb 09 14:24:11 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:11.564+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 35865"
Feb 09 14:24:11 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:11.814+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 44537"
Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.022+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --model /usr/share/ollama/.ollama/models/blobs/sha256-e9758e589d443f653821b7be9bb9092c1bf7434522b70ec6e83591b1320fdb4d --port 39773"
Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.022+01:00 level=INFO source=sched.go:463 msg="system memory" total="31.2 GiB" free="24.1 GiB" free_swap="8.0 GiB"
Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.022+01:00 level=INFO source=sched.go:470 msg="gpu memory" id=GPU-8eabad4d-c702-d0d0-a065-be5c5f83cb97 library=CUDA available="11.8 GiB" free="12.3 GiB" minimum="457.0 MiB" overhead="0 B"
Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.022+01:00 level=INFO source=server.go:757 msg="loading model" "model layers"=37 requested=-1
Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.033+01:00 level=INFO source=runner.go:1411 msg="starting ollama engine"
Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.034+01:00 level=INFO source=runner.go:1446 msg="Server listening on 127.0.0.1:39773"
Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.045+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:4096 KvCacheType: NumThreads:24 GPULayers:37[ID:GPU-8eabad4d-c702-d0d0-a065-be5c5f83cb97 Layers:37(0..36)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.064+01:00 level=INFO source=ggml.go:136 msg="" architecture=qwen25vl file_type=Q4_K_M name="" description="" num_tensors=953 num_key_values=36
Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: load_backend: loaded CPU backend from /usr/local/lib/ollama/libggml-cpu-alderlake.so
Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: ggml_cuda_init: GGML_CUDA_FORCE_MMQ:    no
Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: ggml_cuda_init: found 1 CUDA devices:
Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]:   Device 0: NVIDIA GeForce RTX 5080, compute capability 12.0, VMM: yes, ID: GPU-8eabad4d-c702-d0d0-a065-be5c5f83cb97

OS

WSL2

GPU

Nvidia

CPU

Intel

Ollama version

0.15.6

Originally created by @megengels on GitHub (Feb 9, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/14170 ### What is the issue? ## Bug: Qwen2.5-VL 3B crashes with HTTP 500 on specific PNG image ### Environment - Ollama: v0.15.6 (latest) - OS: Ubuntu 24 (WSL2) - GPU: NVIDIA GeForce RTX 5080 (16GB VRAM) - Model: qwen2.5vl:3b ### Description Qwen2.5-VL 3B consistently crashes with a 500 error on a specific PNG image while a different image with the exact same dimensions works fine. The runner enters a crash-loop after the failure. ### Reproduction 1. Download the two test images attached below (synthetic payslips, fake data) 2. Run the following for each image: ```python import base64, json, urllib.request with open('IMAGE_PATH', 'rb') as f: img_b64 = base64.b64encode(f.read()).decode() data = json.dumps({ 'model': 'qwen2.5vl:3b', 'messages': [{'role': 'user', 'content': 'Quel est le net à payer ?', 'images': [img_b64]}], 'stream': False }).encode() req = urllib.request.Request('http://localhost:11434/api/chat', data=data, headers={'Content-Type': 'application/json'}) resp = urllib.request.urlopen(req, timeout=60) print(resp.read().decode()) ``` ### Results | Image | Size | Dimensions | Result | |-------|------|------------|--------| | bulletin_LÉVY (modern layout) | 7.15 MB | 1654x2339 | ✅ Works | | bulletin_CHARRIER (classic layout) | 4.62 MB | 1654x2339 | ❌ 500 crash | ### Observed behavior - After ~8.5 seconds, the runner crashes - Ollama enters a crash-loop, restarting the runner ~20 times: ``` [GIN] 500 | 8.496894644s | POST "/api/chat" starting runner cmd="/usr/local/bin/ollama runner --ollama-engine --port 34119" starting runner cmd="/usr/local/bin/ollama runner --ollama-engine --port 35553" starting runner cmd="/usr/local/bin/ollama runner --ollama-engine --port 45419" ... (~20 more restarts) ``` ### Additional context - Both images are the same dimensions (1654x2339), crashing one is actually smaller (4.62 MB vs 7.15 MB) - Plenty of VRAM available (12+ GB free / 16 GB total) - No other processes running on GPU - Reproduced consistently across multiple attempts and service restarts - Test images are synthetic French payslips (fake data), safe to attach <img width="1654" height="2339" alt="Image" src="https://github.com/user-attachments/assets/fc632dce-28f5-4cfb-acf8-9d4566b42130" /> <img width="1654" height="2339" alt="Image" src="https://github.com/user-attachments/assets/bd3133e5-294d-411d-b6d3-8f4dad8db3d9" /> ### Relevant log output ```shell Feb 09 14:18:57 MEGDIG-D-MES ollama[605557]: time=2026-02-09T14:18:57.793+01:00 level=INFO source=runner.go:464 msg="failure during GPU discovery" OLLAMA_LIBRARY_PATH="[/usr/local/lib/ollama /usr/local/lib/ollama/cuda_v13]" extra_envs=map[] error="failed to finish discovery before timeout" Feb 09 14:18:57 MEGDIG-D-MES ollama[605557]: time=2026-02-09T14:18:57.793+01:00 level=WARN source=runner.go:356 msg="unable to refresh free memory, using old values" Feb 09 14:23:03 MEGDIG-D-MES ollama[654807]: [GIN] 2026/02/09 - 14:23:03 | 500 | 8.496894644s | 127.0.0.1 | POST "/api/chat" Feb 09 14:24:06 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:06.604+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 34119" Feb 09 14:24:07 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:07.069+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 35553" Feb 09 14:24:07 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:07.314+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 45419" Feb 09 14:24:07 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:07.564+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 44979" Feb 09 14:24:07 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:07.814+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 37187" Feb 09 14:24:08 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:08.064+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 45377" Feb 09 14:24:08 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:08.314+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 33263" Feb 09 14:24:08 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:08.564+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 38835" Feb 09 14:24:08 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:08.814+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 42785" Feb 09 14:24:09 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:09.064+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 38421" Feb 09 14:24:09 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:09.314+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 36565" Feb 09 14:24:09 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:09.564+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 34367" Feb 09 14:24:09 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:09.814+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 40257" Feb 09 14:24:10 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:10.064+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 44767" Feb 09 14:24:10 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:10.314+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 45329" Feb 09 14:24:10 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:10.564+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 34121" Feb 09 14:24:10 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:10.814+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 36567" Feb 09 14:24:11 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:11.065+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 35639" Feb 09 14:24:11 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:11.314+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 36387" Feb 09 14:24:11 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:11.564+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 35865" Feb 09 14:24:11 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:11.814+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 44537" Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.022+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --model /usr/share/ollama/.ollama/models/blobs/sha256-e9758e589d443f653821b7be9bb9092c1bf7434522b70ec6e83591b1320fdb4d --port 39773" Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.022+01:00 level=INFO source=sched.go:463 msg="system memory" total="31.2 GiB" free="24.1 GiB" free_swap="8.0 GiB" Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.022+01:00 level=INFO source=sched.go:470 msg="gpu memory" id=GPU-8eabad4d-c702-d0d0-a065-be5c5f83cb97 library=CUDA available="11.8 GiB" free="12.3 GiB" minimum="457.0 MiB" overhead="0 B" Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.022+01:00 level=INFO source=server.go:757 msg="loading model" "model layers"=37 requested=-1 Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.033+01:00 level=INFO source=runner.go:1411 msg="starting ollama engine" Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.034+01:00 level=INFO source=runner.go:1446 msg="Server listening on 127.0.0.1:39773" Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.045+01:00 level=INFO source=runner.go:1284 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Disabled KvSize:4096 KvCacheType: NumThreads:24 GPULayers:37[ID:GPU-8eabad4d-c702-d0d0-a065-be5c5f83cb97 Layers:37(0..36)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: time=2026-02-09T14:24:12.064+01:00 level=INFO source=ggml.go:136 msg="" architecture=qwen25vl file_type=Q4_K_M name="" description="" num_tensors=953 num_key_values=36 Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: load_backend: loaded CPU backend from /usr/local/lib/ollama/libggml-cpu-alderlake.so Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: ggml_cuda_init: found 1 CUDA devices: Feb 09 14:24:12 MEGDIG-D-MES ollama[654807]: Device 0: NVIDIA GeForce RTX 5080, compute capability 12.0, VMM: yes, ID: GPU-8eabad4d-c702-d0d0-a065-be5c5f83cb97 ``` ### OS WSL2 ### GPU Nvidia ### CPU Intel ### Ollama version 0.15.6
GiteaMirror added the bug label 2026-04-29 09:41:34 -05:00
GiteaMirror commented 2026-04-29 09:41:37 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 9, 2026):

Ollama is not crash-looping, it's probing the GPU devices because earlier on it failed to do proper dicvoery (failure during GPU discovery).

The crash is because the runner was unable to extract a token from the sampler. It's not clear why, but it is reproducible:

ollama  | panic: failed to sample token
ollama  | 
ollama  | goroutine 1250 [running]:
ollama  | github.com/ollama/ollama/runner/ollamarunner.(*Server).computeBatch(0xc0000374a0, {0x14, {0x5a84bf4238b0, 0xc000478000}, {0x5a84bf430340, 0xc0020a97d0}, {0xc000fb8080, 0xd, 0x10}, {{0x5a84bf430340, ...}, ...}, ...})
ollama  | 	github.com/ollama/ollama/runner/ollamarunner/runner.go:762 +0x1c25
ollama  | created by github.com/ollama/ollama/runner/ollamarunner.(*Server).run in goroutine 67
ollama  | 	github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x2cd
ollama  | time=2026-02-09T15:23:37.947Z level=ERROR source=server.go:1610 msg="post predict" error="Post \"http://127.0.0.1:42547/completion\": EOF"
ollama  | [GIN] 2026/02/09 - 15:23:37 | 500 |  3.176837401s |      172.18.0.1 | POST     "/api/generate"
<!-- gh-comment-id:3872450025 --> @rick-github commented on GitHub (Feb 9, 2026): Ollama is not crash-looping, it's probing the GPU devices because earlier on it failed to do proper dicvoery (`failure during GPU discovery`). The crash is because the runner was unable to extract a token from the sampler. It's not clear why, but it is reproducible: ``` ollama | panic: failed to sample token ollama | ollama | goroutine 1250 [running]: ollama | github.com/ollama/ollama/runner/ollamarunner.(*Server).computeBatch(0xc0000374a0, {0x14, {0x5a84bf4238b0, 0xc000478000}, {0x5a84bf430340, 0xc0020a97d0}, {0xc000fb8080, 0xd, 0x10}, {{0x5a84bf430340, ...}, ...}, ...}) ollama | github.com/ollama/ollama/runner/ollamarunner/runner.go:762 +0x1c25 ollama | created by github.com/ollama/ollama/runner/ollamarunner.(*Server).run in goroutine 67 ollama | github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x2cd ollama | time=2026-02-09T15:23:37.947Z level=ERROR source=server.go:1610 msg="post predict" error="Post \"http://127.0.0.1:42547/completion\": EOF" ollama | [GIN] 2026/02/09 - 15:23:37 | 500 | 3.176837401s | 172.18.0.1 | POST "/api/generate" ```
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#55750
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?