github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #14110] 0.15.5-rc0 GPU detection regression #34966

New Issue
Closed
opened 2026-04-22 19:03:13 -05:00 by GiteaMirror · 5 comments
GiteaMirror commented 2026-04-22 19:03:13 -05:00
Owner
Copy Link

Originally created by @Vyryn on GitHub (Feb 6, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/14110

What is the issue?

Hello. I am running the 0.15.5-rc0 prerelease on fedora 43 desktop (no docker).
ollama uses my cpu instead of gpu (ollama ps shows 100% CPU). This persists when I manually set CUDA_VISIBLE_DEVICES=[guid], OLLAMA_LLM_LIBRARY=cuda_v11 and/or enable vulkan.

When I downgrade to 0.15.4 release and make no other changes, my GPU is successfully detected and used.

Relevant log output

> nvidia-smi

Thu Feb  5 21:38:04 2026
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.119.02             Driver Version: 580.119.02     CUDA Version: 13.0     |
+-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA GeForce RTX 3070 Ti     Off |   00000000:01:00.0  On |                  N/A |
|  0%   47C    P3             48W /  310W |    1260MiB /   8192MiB |     15%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

> nvidia-smi -L
> GPU 0: NVIDIA GeForce RTX 3070 Ti (UUID: GPU-XXXXXXXX)


When I launch with none of the troubleshooting env variables set:

Feb 05 21:29:22 fedora systemd[1]: Started ollama.service - Ollama Service.
Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.215-05:00 level=INFO source=routes.go:1622 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GF>
Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.216-05:00 level=INFO source=images.go:473 msg="total blobs: 8"
Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.216-05:00 level=INFO source=images.go:480 msg="total unused blobs removed: 0"
Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.216-05:00 level=INFO source=routes.go:1675 msg="Listening on 127.0.0.1:11434 (version 0.15.5-rc0)"
Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.216-05:00 level=DEBUG source=sched.go:121 msg="starting llm scheduler"
Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.222-05:00 level=INFO source=runner.go:67 msg="discovering available GPUs..."
Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.222-05:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 35495"
Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.222-05:00 level=DEBUG source=server.go:431 msg=subprocess PATH=/usr/bin OLLAMA_DEBUG=1 LD_LIBRARY_PATH=/usr/bin OLLAMA_LIBRARY_PATH=/usr/bin
Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.252-05:00 level=DEBUG source=runner.go:437 msg="bootstrap discovery took" duration=30.225075ms OLLAMA_LIBRARY_PATH=[/usr/bin] extra_envs=map[]
Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.252-05:00 level=DEBUG source=runner.go:124 msg="evaluating which, if any, devices to filter out" initial_count=0
Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.252-05:00 level=DEBUG source=runner.go:40 msg="GPU bootstrap discovery took" duration=35.801358ms
Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.252-05:00 level=INFO source=types.go:60 msg="inference compute" id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="31.1 G>
Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.252-05:00 level=INFO source=routes.go:1725 msg="vram-based default context" total_vram="0 B" default_num_ctx=4096

When I launch with CUDA_VISIBLE_DEVICES set: (also tried with gpu id, 0)

Feb 05 21:26:36 fedora systemd[1]: Started ollama.service - Ollama Service.
Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.247-05:00 level=INFO source=routes.go:1622 msg="server config" env="map[CUDA_VISIBLE_DEVICES:GPU-XXXXXXXX GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDIN>
Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.248-05:00 level=INFO source=images.go:473 msg="total blobs: 8"
Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.248-05:00 level=INFO source=images.go:480 msg="total unused blobs removed: 0"
Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.248-05:00 level=INFO source=routes.go:1675 msg="Listening on 127.0.0.1:11434 (version 0.15.5-rc0)"
Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.248-05:00 level=DEBUG source=sched.go:121 msg="starting llm scheduler"
Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.255-05:00 level=INFO source=runner.go:67 msg="discovering available GPUs..."
Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.255-05:00 level=WARN source=runner.go:485 msg="user overrode visible devices" CUDA_VISIBLE_DEVICES=GPU-XXXXXXXX
Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.255-05:00 level=WARN source=runner.go:489 msg="if GPUs are not correctly discovered, unset and try again"
Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.256-05:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 39353"
Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.256-05:00 level=DEBUG source=server.go:431 msg=subprocess PATH=/usr/bin OLLAMA_DEBUG=1 CUDA_VISIBLE_DEVICES=GPU-XXXXXXXX LD_LIBRARY_PATH=/usr/bin O>
Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.286-05:00 level=DEBUG source=runner.go:437 msg="bootstrap discovery took" duration=30.68834ms OLLAMA_LIBRARY_PATH=[/usr/bin] extra_envs=map[]
Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.286-05:00 level=DEBUG source=runner.go:124 msg="evaluating which, if any, devices to filter out" initial_count=0
Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.286-05:00 level=DEBUG source=runner.go:40 msg="GPU bootstrap discovery took" duration=37.715295ms
Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.286-05:00 level=INFO source=types.go:60 msg="inference compute" id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="31.1 G>
Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.286-05:00 level=INFO source=routes.go:1725 msg="vram-based default context" total_vram="0 B" default_num_ctx=4096


When I launch with OLLAMA_LLM_LIBRARY=cuda_v11:

Feb 05 21:34:41 fedora systemd[1]: Started ollama.service - Ollama Service.
Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.443-05:00 level=INFO source=routes.go:1622 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GF>
Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.444-05:00 level=INFO source=images.go:473 msg="total blobs: 8"
Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.445-05:00 level=INFO source=images.go:480 msg="total unused blobs removed: 0"
Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.445-05:00 level=INFO source=routes.go:1675 msg="Listening on 127.0.0.1:11434 (version 0.15.5-rc0)"
Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.445-05:00 level=DEBUG source=sched.go:121 msg="starting llm scheduler"
Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.451-05:00 level=INFO source=runner.go:67 msg="discovering available GPUs..."
Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.452-05:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 38775"
Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.452-05:00 level=DEBUG source=server.go:431 msg=subprocess PATH=/usr/bin OLLAMA_DEBUG=1 OLLAMA_LLM_LIBRARY=cuda_v11 LD_LIBRARY_PATH=/usr/bin OLLAMA_LIBRARY_PATH=/usr/bin
Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.481-05:00 level=DEBUG source=runner.go:437 msg="bootstrap discovery took" duration=30.148575ms OLLAMA_LIBRARY_PATH=[/usr/bin] extra_envs=map[]
Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.481-05:00 level=DEBUG source=runner.go:124 msg="evaluating which, if any, devices to filter out" initial_count=0
Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.481-05:00 level=DEBUG source=runner.go:40 msg="GPU bootstrap discovery took" duration=36.495709ms
Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.482-05:00 level=INFO source=types.go:60 msg="inference compute" id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="31.1 G>
Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.482-05:00 level=INFO source=routes.go:1725 msg="vram-based default context" total_vram="0 B" default_num_ctx=4096


I also tried the combination of  CUDA_VISIBLE_DEVICES and OLLAMA_LLM_LIBRARY, and I also tried vulkan

ollama.service contents:

[Unit]
Description=Ollama Service
After=network-online.target

[Service]
ExecStart=/usr/bin/ollama serve
User=ollama
Group=ollama
Restart=always
RestartSec=3
Environment="PATH=/usr/bin"
Environment="HOME=/usr/share/ollama"
Environment="OLLAMA_DEBUG=1"
# Environment="CUDA_VISIBLE_DEVICES=GPU-XXXXXXXX"
# Environment="OLLAMA_LLM_LIBRARY=cuda_v11"
# Environment="OLLAMA_VULKAN=1"


[Install]
WantedBy=default.target


Downgrading to ollama 0.15.4 release version and re-running causes my GPU to be detected and successfully run on:

Feb 05 21:48:26 fedora systemd[1]: Started ollama.service - Ollama Service.
Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.098-05:00 level=INFO source=routes.go:1631 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GF>
Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.099-05:00 level=INFO source=images.go:473 msg="total blobs: 8"
Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.099-05:00 level=INFO source=images.go:480 msg="total unused blobs removed: 0"
Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.099-05:00 level=INFO source=routes.go:1684 msg="Listening on 127.0.0.1:11434 (version 0.15.4)"
Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.099-05:00 level=INFO source=runner.go:67 msg="discovering available GPUs..."
Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.099-05:00 level=INFO source=runner.go:106 msg="experimental Vulkan support disabled.  To enable, set OLLAMA_VULKAN=1"
Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.100-05:00 level=INFO source=server.go:429 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 38443"
Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.212-05:00 level=INFO source=server.go:429 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 44527"
Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.309-05:00 level=INFO source=server.go:429 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 33985"
Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.309-05:00 level=INFO source=server.go:429 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 43453"
Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.429-05:00 level=INFO source=types.go:42 msg="inference compute" id=GPU-XXXXXXXX filter_id="" library=CUDA compute=8.6 name=CUDA0 description="NVIDI>
Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.429-05:00 level=INFO source=routes.go:1725 msg="entering low vram mode" "total vram"="8.0 GiB" threshold="20.0 GiB"

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.15.5-rc0

Originally created by @Vyryn on GitHub (Feb 6, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/14110 ### What is the issue? Hello. I am running the 0.15.5-rc0 prerelease on fedora 43 desktop (no docker). ollama uses my cpu instead of gpu (ollama ps shows 100% CPU). This persists when I manually set CUDA_VISIBLE_DEVICES=[guid], OLLAMA_LLM_LIBRARY=cuda_v11 and/or enable vulkan. When I downgrade to 0.15.4 release and make no other changes, my GPU is successfully detected and used. ### Relevant log output ```shell > nvidia-smi Thu Feb 5 21:38:04 2026 +-----------------------------------------------------------------------------------------+ | NVIDIA-SMI 580.119.02 Driver Version: 580.119.02 CUDA Version: 13.0 | +-----------------------------------------+------------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+========================+======================| | 0 NVIDIA GeForce RTX 3070 Ti Off | 00000000:01:00.0 On | N/A | | 0% 47C P3 48W / 310W | 1260MiB / 8192MiB | 15% Default | | | | N/A | +-----------------------------------------+------------------------+----------------------+ > nvidia-smi -L > GPU 0: NVIDIA GeForce RTX 3070 Ti (UUID: GPU-XXXXXXXX) When I launch with none of the troubleshooting env variables set: Feb 05 21:29:22 fedora systemd[1]: Started ollama.service - Ollama Service. Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.215-05:00 level=INFO source=routes.go:1622 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GF> Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.216-05:00 level=INFO source=images.go:473 msg="total blobs: 8" Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.216-05:00 level=INFO source=images.go:480 msg="total unused blobs removed: 0" Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.216-05:00 level=INFO source=routes.go:1675 msg="Listening on 127.0.0.1:11434 (version 0.15.5-rc0)" Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.216-05:00 level=DEBUG source=sched.go:121 msg="starting llm scheduler" Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.222-05:00 level=INFO source=runner.go:67 msg="discovering available GPUs..." Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.222-05:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 35495" Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.222-05:00 level=DEBUG source=server.go:431 msg=subprocess PATH=/usr/bin OLLAMA_DEBUG=1 LD_LIBRARY_PATH=/usr/bin OLLAMA_LIBRARY_PATH=/usr/bin Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.252-05:00 level=DEBUG source=runner.go:437 msg="bootstrap discovery took" duration=30.225075ms OLLAMA_LIBRARY_PATH=[/usr/bin] extra_envs=map[] Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.252-05:00 level=DEBUG source=runner.go:124 msg="evaluating which, if any, devices to filter out" initial_count=0 Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.252-05:00 level=DEBUG source=runner.go:40 msg="GPU bootstrap discovery took" duration=35.801358ms Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.252-05:00 level=INFO source=types.go:60 msg="inference compute" id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="31.1 G> Feb 05 21:29:22 fedora ollama[23390]: time=2026-02-05T21:29:22.252-05:00 level=INFO source=routes.go:1725 msg="vram-based default context" total_vram="0 B" default_num_ctx=4096 When I launch with CUDA_VISIBLE_DEVICES set: (also tried with gpu id, 0) Feb 05 21:26:36 fedora systemd[1]: Started ollama.service - Ollama Service. Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.247-05:00 level=INFO source=routes.go:1622 msg="server config" env="map[CUDA_VISIBLE_DEVICES:GPU-XXXXXXXX GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDIN> Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.248-05:00 level=INFO source=images.go:473 msg="total blobs: 8" Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.248-05:00 level=INFO source=images.go:480 msg="total unused blobs removed: 0" Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.248-05:00 level=INFO source=routes.go:1675 msg="Listening on 127.0.0.1:11434 (version 0.15.5-rc0)" Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.248-05:00 level=DEBUG source=sched.go:121 msg="starting llm scheduler" Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.255-05:00 level=INFO source=runner.go:67 msg="discovering available GPUs..." Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.255-05:00 level=WARN source=runner.go:485 msg="user overrode visible devices" CUDA_VISIBLE_DEVICES=GPU-XXXXXXXX Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.255-05:00 level=WARN source=runner.go:489 msg="if GPUs are not correctly discovered, unset and try again" Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.256-05:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 39353" Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.256-05:00 level=DEBUG source=server.go:431 msg=subprocess PATH=/usr/bin OLLAMA_DEBUG=1 CUDA_VISIBLE_DEVICES=GPU-XXXXXXXX LD_LIBRARY_PATH=/usr/bin O> Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.286-05:00 level=DEBUG source=runner.go:437 msg="bootstrap discovery took" duration=30.68834ms OLLAMA_LIBRARY_PATH=[/usr/bin] extra_envs=map[] Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.286-05:00 level=DEBUG source=runner.go:124 msg="evaluating which, if any, devices to filter out" initial_count=0 Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.286-05:00 level=DEBUG source=runner.go:40 msg="GPU bootstrap discovery took" duration=37.715295ms Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.286-05:00 level=INFO source=types.go:60 msg="inference compute" id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="31.1 G> Feb 05 21:26:36 fedora ollama[22570]: time=2026-02-05T21:26:36.286-05:00 level=INFO source=routes.go:1725 msg="vram-based default context" total_vram="0 B" default_num_ctx=4096 When I launch with OLLAMA_LLM_LIBRARY=cuda_v11: Feb 05 21:34:41 fedora systemd[1]: Started ollama.service - Ollama Service. Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.443-05:00 level=INFO source=routes.go:1622 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GF> Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.444-05:00 level=INFO source=images.go:473 msg="total blobs: 8" Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.445-05:00 level=INFO source=images.go:480 msg="total unused blobs removed: 0" Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.445-05:00 level=INFO source=routes.go:1675 msg="Listening on 127.0.0.1:11434 (version 0.15.5-rc0)" Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.445-05:00 level=DEBUG source=sched.go:121 msg="starting llm scheduler" Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.451-05:00 level=INFO source=runner.go:67 msg="discovering available GPUs..." Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.452-05:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 38775" Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.452-05:00 level=DEBUG source=server.go:431 msg=subprocess PATH=/usr/bin OLLAMA_DEBUG=1 OLLAMA_LLM_LIBRARY=cuda_v11 LD_LIBRARY_PATH=/usr/bin OLLAMA_LIBRARY_PATH=/usr/bin Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.481-05:00 level=DEBUG source=runner.go:437 msg="bootstrap discovery took" duration=30.148575ms OLLAMA_LIBRARY_PATH=[/usr/bin] extra_envs=map[] Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.481-05:00 level=DEBUG source=runner.go:124 msg="evaluating which, if any, devices to filter out" initial_count=0 Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.481-05:00 level=DEBUG source=runner.go:40 msg="GPU bootstrap discovery took" duration=36.495709ms Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.482-05:00 level=INFO source=types.go:60 msg="inference compute" id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="31.1 G> Feb 05 21:34:41 fedora ollama[24671]: time=2026-02-05T21:34:41.482-05:00 level=INFO source=routes.go:1725 msg="vram-based default context" total_vram="0 B" default_num_ctx=4096 I also tried the combination of CUDA_VISIBLE_DEVICES and OLLAMA_LLM_LIBRARY, and I also tried vulkan ollama.service contents: [Unit] Description=Ollama Service After=network-online.target [Service] ExecStart=/usr/bin/ollama serve User=ollama Group=ollama Restart=always RestartSec=3 Environment="PATH=/usr/bin" Environment="HOME=/usr/share/ollama" Environment="OLLAMA_DEBUG=1" # Environment="CUDA_VISIBLE_DEVICES=GPU-XXXXXXXX" # Environment="OLLAMA_LLM_LIBRARY=cuda_v11" # Environment="OLLAMA_VULKAN=1" [Install] WantedBy=default.target Downgrading to ollama 0.15.4 release version and re-running causes my GPU to be detected and successfully run on: Feb 05 21:48:26 fedora systemd[1]: Started ollama.service - Ollama Service. Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.098-05:00 level=INFO source=routes.go:1631 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GF> Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.099-05:00 level=INFO source=images.go:473 msg="total blobs: 8" Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.099-05:00 level=INFO source=images.go:480 msg="total unused blobs removed: 0" Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.099-05:00 level=INFO source=routes.go:1684 msg="Listening on 127.0.0.1:11434 (version 0.15.4)" Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.099-05:00 level=INFO source=runner.go:67 msg="discovering available GPUs..." Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.099-05:00 level=INFO source=runner.go:106 msg="experimental Vulkan support disabled. To enable, set OLLAMA_VULKAN=1" Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.100-05:00 level=INFO source=server.go:429 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 38443" Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.212-05:00 level=INFO source=server.go:429 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 44527" Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.309-05:00 level=INFO source=server.go:429 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 33985" Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.309-05:00 level=INFO source=server.go:429 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 43453" Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.429-05:00 level=INFO source=types.go:42 msg="inference compute" id=GPU-XXXXXXXX filter_id="" library=CUDA compute=8.6 name=CUDA0 description="NVIDI> Feb 05 21:48:26 fedora ollama[27889]: time=2026-02-05T21:48:26.429-05:00 level=INFO source=routes.go:1725 msg="entering low vram mode" "total vram"="8.0 GiB" threshold="20.0 GiB" ``` ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.15.5-rc0
GiteaMirror added the bug label 2026-04-22 19:03:13 -05:00
GiteaMirror commented 2026-04-22 19:03:15 -05:00
Author
Owner
Copy Link

@MichalRIcar commented on GitHub (Feb 6, 2026):

+1 for AMD card with the latest drivers rocm - running on just released 0.15.5 - the exact same behavior as above

<!-- gh-comment-id:3859345544 --> @MichalRIcar commented on GitHub (Feb 6, 2026): +1 for AMD card with the latest drivers rocm - running on just released 0.15.5 - the exact same behavior as above
GiteaMirror commented 2026-04-22 19:03:15 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 6, 2026):

Set OLLAMA_DEBUG=2 and post the log from the start through to the line that contains inference compute.

<!-- gh-comment-id:3859502049 --> @rick-github commented on GitHub (Feb 6, 2026): Set `OLLAMA_DEBUG=2` and post the log from the start through to the line that contains `inference compute`.
GiteaMirror commented 2026-04-22 19:03:16 -05:00
Author
Owner
Copy Link

@MichalRIcar commented on GitHub (Feb 6, 2026):

**ollama 0.15.4 - speed 30t/s >>>**
Feb 06 11:59:19  ollama[219345]: load_backend: loaded ROCm backend from /usr/local/lib/ollama/rocm/libggml-hip.so
Feb 06 11:59:20  ollama[219345]: time=2026-02-06T11:59:20.179+01:00 level=INFO source=runner.go:1278 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:8192 KvCacheType: NumThreads:32 G>
Feb 06 11:59:20  ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=runner.go:1278 msg=load request="{Operation:commit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:8192 KvCacheType: NumThreads:32 >
Feb 06 11:59:20  ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=ggml.go:482 msg="offloading 64 repeating layers to GPU"
Feb 06 11:59:20  ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=ggml.go:489 msg="offloading output layer to GPU"
Feb 06 11:59:20  ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=ggml.go:494 msg="offloaded 65/65 layers to GPU"
Feb 06 11:59:20  ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=device.go:240 msg="model weights" device=ROCm0 size="19.1 GiB"
Feb 06 11:59:20  ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=device.go:245 msg="model weights" device=CPU size="417.3 MiB"
Feb 06 11:59:20  ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=device.go:251 msg="kv cache" device=ROCm0 size="2.0 GiB"
Feb 06 11:59:20  ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=device.go:262 msg="compute graph" device=ROCm0 size="1000.6 MiB"
Feb 06 11:59:20  ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=device.go:267 msg="compute graph" device=CPU size="79.1 MiB"
Feb 06 11:59:20  ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=device.go:272 msg="total memory" size="22.5 GiB"
Feb 06 11:59:20  ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=sched.go:526 msg="loaded runners" count=1
Feb 06 11:59:20  ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=server.go:1347 msg="waiting for llama runner to start responding"
Feb 06 11:59:20  ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=server.go:1381 msg="waiting for server to become available" status="llm server loading model"
Feb 06 11:59:23  ollama[219345]: time=2026-02-06T11:59:23.119+01:00 level=INFO source=server.go:1385 msg="llama runner started in 4.65 seconds"

**ollama 0.15.5 - speed 8t/s >>>**
time=2026-02-06T11:36:10.103+01:00 level=INFO source=runner.go:1283 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:32768 KvCacheType: NumThreads:32 GP>
time=2026-02-06T11:36:10.433+01:00 level=INFO source=runner.go:1283 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:32768 KvCacheType: NumThreads:32 >
time=2026-02-06T11:36:11.376+01:00 level=INFO source=runner.go:1283 msg=load request="{Operation:commit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:32768 KvCacheType: NumThreads:32>
time=2026-02-06T11:36:11.376+01:00 level=INFO source=ggml.go:482 msg="offloading 55 repeating layers to GPU"
time=2026-02-06T11:36:11.376+01:00 level=INFO source=ggml.go:486 msg="offloading output layer to CPU"
time=2026-02-06T11:36:11.376+01:00 level=INFO source=ggml.go:494 msg="offloaded 55/65 layers to GPU"
time=2026-02-06T11:36:11.376+01:00 level=INFO source=device.go:240 msg="model weights" device=ROCm0 size="14.8 GiB"
time=2026-02-06T11:36:11.376+01:00 level=INFO source=device.go:245 msg="model weights" device=CPU size="4.6 GiB"
time=2026-02-06T11:36:11.376+01:00 level=INFO source=device.go:251 msg="kv cache" device=ROCm0 size="6.9 GiB"
time=2026-02-06T11:36:11.376+01:00 level=INFO source=device.go:256 msg="kv cache" device=CPU size="1.1 GiB"
time=2026-02-06T11:36:11.376+01:00 level=INFO source=device.go:262 msg="compute graph" device=ROCm0 size="1.2 GiB"
time=2026-02-06T11:36:11.376+01:00 level=INFO source=device.go:267 msg="compute graph" device=CPU size="142.4 MiB"
time=2026-02-06T11:36:11.376+01:00 level=INFO source=device.go:272 msg="total memory" size="28.8 GiB"
time=2026-02-06T11:36:11.376+01:00 level=INFO source=sched.go:537 msg="loaded runners" count=1
time=2026-02-06T11:36:11.376+01:00 level=INFO source=server.go:1349 msg="waiting for llama runner to start responding"
time=2026-02-06T11:36:11.376+01:00 level=INFO source=server.go:1383 msg="waiting for server to become available" status="llm server loading model"
time=2026-02-06T11:36:13.637+01:00 level=INFO source=server.go:1387 msg="llama runner started in 5.34 seconds"
<!-- gh-comment-id:3859747556 --> @MichalRIcar commented on GitHub (Feb 6, 2026): ``` **ollama 0.15.4 - speed 30t/s >>>** Feb 06 11:59:19 ollama[219345]: load_backend: loaded ROCm backend from /usr/local/lib/ollama/rocm/libggml-hip.so Feb 06 11:59:20 ollama[219345]: time=2026-02-06T11:59:20.179+01:00 level=INFO source=runner.go:1278 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:8192 KvCacheType: NumThreads:32 G> Feb 06 11:59:20 ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=runner.go:1278 msg=load request="{Operation:commit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:8192 KvCacheType: NumThreads:32 > Feb 06 11:59:20 ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=ggml.go:482 msg="offloading 64 repeating layers to GPU" Feb 06 11:59:20 ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=ggml.go:489 msg="offloading output layer to GPU" Feb 06 11:59:20 ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=ggml.go:494 msg="offloaded 65/65 layers to GPU" Feb 06 11:59:20 ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=device.go:240 msg="model weights" device=ROCm0 size="19.1 GiB" Feb 06 11:59:20 ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=device.go:245 msg="model weights" device=CPU size="417.3 MiB" Feb 06 11:59:20 ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=device.go:251 msg="kv cache" device=ROCm0 size="2.0 GiB" Feb 06 11:59:20 ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=device.go:262 msg="compute graph" device=ROCm0 size="1000.6 MiB" Feb 06 11:59:20 ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=device.go:267 msg="compute graph" device=CPU size="79.1 MiB" Feb 06 11:59:20 ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=device.go:272 msg="total memory" size="22.5 GiB" Feb 06 11:59:20 ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=sched.go:526 msg="loaded runners" count=1 Feb 06 11:59:20 ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=server.go:1347 msg="waiting for llama runner to start responding" Feb 06 11:59:20 ollama[219345]: time=2026-02-06T11:59:20.605+01:00 level=INFO source=server.go:1381 msg="waiting for server to become available" status="llm server loading model" Feb 06 11:59:23 ollama[219345]: time=2026-02-06T11:59:23.119+01:00 level=INFO source=server.go:1385 msg="llama runner started in 4.65 seconds" **ollama 0.15.5 - speed 8t/s >>>** time=2026-02-06T11:36:10.103+01:00 level=INFO source=runner.go:1283 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:32768 KvCacheType: NumThreads:32 GP> time=2026-02-06T11:36:10.433+01:00 level=INFO source=runner.go:1283 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:32768 KvCacheType: NumThreads:32 > time=2026-02-06T11:36:11.376+01:00 level=INFO source=runner.go:1283 msg=load request="{Operation:commit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:32768 KvCacheType: NumThreads:32> time=2026-02-06T11:36:11.376+01:00 level=INFO source=ggml.go:482 msg="offloading 55 repeating layers to GPU" time=2026-02-06T11:36:11.376+01:00 level=INFO source=ggml.go:486 msg="offloading output layer to CPU" time=2026-02-06T11:36:11.376+01:00 level=INFO source=ggml.go:494 msg="offloaded 55/65 layers to GPU" time=2026-02-06T11:36:11.376+01:00 level=INFO source=device.go:240 msg="model weights" device=ROCm0 size="14.8 GiB" time=2026-02-06T11:36:11.376+01:00 level=INFO source=device.go:245 msg="model weights" device=CPU size="4.6 GiB" time=2026-02-06T11:36:11.376+01:00 level=INFO source=device.go:251 msg="kv cache" device=ROCm0 size="6.9 GiB" time=2026-02-06T11:36:11.376+01:00 level=INFO source=device.go:256 msg="kv cache" device=CPU size="1.1 GiB" time=2026-02-06T11:36:11.376+01:00 level=INFO source=device.go:262 msg="compute graph" device=ROCm0 size="1.2 GiB" time=2026-02-06T11:36:11.376+01:00 level=INFO source=device.go:267 msg="compute graph" device=CPU size="142.4 MiB" time=2026-02-06T11:36:11.376+01:00 level=INFO source=device.go:272 msg="total memory" size="28.8 GiB" time=2026-02-06T11:36:11.376+01:00 level=INFO source=sched.go:537 msg="loaded runners" count=1 time=2026-02-06T11:36:11.376+01:00 level=INFO source=server.go:1349 msg="waiting for llama runner to start responding" time=2026-02-06T11:36:11.376+01:00 level=INFO source=server.go:1383 msg="waiting for server to become available" status="llm server loading model" time=2026-02-06T11:36:13.637+01:00 level=INFO source=server.go:1387 msg="llama runner started in 5.34 seconds" ```
GiteaMirror commented 2026-04-22 19:03:17 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 6, 2026):

@MichalRIcar Your problem is unrelated, it is this one: #14116.

<!-- gh-comment-id:3859809021 --> @rick-github commented on GitHub (Feb 6, 2026): @MichalRIcar Your problem is unrelated, it is this one: #14116.
GiteaMirror commented 2026-04-22 19:03:19 -05:00
Author
Owner
Copy Link

@Vyryn commented on GitHub (Feb 6, 2026):

This issue is not present for me on the 0.15.5 release version, so it's probably not worth investigating further

<!-- gh-comment-id:3861043428 --> @Vyryn commented on GitHub (Feb 6, 2026): This issue is not present for me on the 0.15.5 release version, so it's probably not worth investigating further
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#34966
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?