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 #15321] Adding cap_perfmon to ollama breaks GPU discovery #71860

New Issue
Open
opened 2026-05-05 02:45:50 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-05-05 02:45:50 -05:00
Owner
Copy Link

Originally created by @relzp on GitHub (Apr 4, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/15321

What is the issue?

Following the docs at https://docs.ollama.com/gpu#vulkan-gpu-support, I added the perfmon capability to the ollama executable using the command sudo setcap cap_perfmon+ep /usr/bin/ollama to give ollama access to the available VRAM data.

However when restarting the ollama service, I noticed that neither of my GPUs were detected. I usually use Vulkan, but I also tested with ROCm, and I noticed the same behaviour (GPU detected without cap_perfmon, not detected with cap_perfmon).

Relevant log output

without cap_perfmon:
Started Ollama Service.
time=2026-04-04T08:08:52.131Z level=INFO source=routes.go:1744 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:0 OLLAMA_DEBUG:INFO OLLAMA_DEBUG_LOG_REQUESTS:false OLLAMA_EDITOR: OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:/var/lib/ollama OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NO_CLOUD:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false OLLAMA_VULKAN:false ROCR_VISIBLE_DEVICES: http_proxy: https_proxy: no_proxy:]"
time=2026-04-04T08:08:52.131Z level=INFO source=routes.go:1746 msg="Ollama cloud disabled: false"
time=2026-04-04T08:08:52.132Z level=INFO source=images.go:499 msg="total blobs: 29"
time=2026-04-04T08:08:52.132Z level=INFO source=images.go:506 msg="total unused blobs removed: 0"
time=2026-04-04T08:08:52.133Z level=INFO source=routes.go:1802 msg="Listening on 127.0.0.1:11434 (version 0.20.0)"
time=2026-04-04T08:08:52.133Z level=INFO source=runner.go:67 msg="discovering available GPUs..."
time=2026-04-04T08:08:52.133Z level=INFO source=server.go:432 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 36717"
time=2026-04-04T08:08:52.268Z level=INFO source=types.go:42 msg="inference compute" id=00000000-0c00-0000-0000-000000000000 filter_id="" library=Vulkan compute=0.0 name=Vulkan0 description="AMD Radeon RX 7700 XT (RADV NAVI32)" libdirs=ollama driver=0.0 pci_id=0000:0c:00.0 type=discrete total="12.0 GiB" available="9.1 GiB"
time=2026-04-04T08:08:52.268Z level=INFO source=types.go:42 msg="inference compute" id=1736597e-c3c7-5b7d-9882-ade1b4a6ea1b filter_id="" library=Vulkan compute=0.0 name=Vulkan1 description="NVIDIA GeForce GTX 1060 6GB" libdirs=ollama driver=0.0 pci_id=0000:0d:00.0 type=discrete total="6.0 GiB" available="5.5 GiB"
time=2026-04-04T08:08:52.268Z level=INFO source=routes.go:1852 msg="vram-based default context" total_vram="18.0 GiB" default_num_ctx=4096


with cap_perfmon:
Started Ollama Service.
time=2026-04-04T08:09:23.746Z level=INFO source=routes.go:1744 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:0 OLLAMA_DEBUG:INFO OLLAMA_DEBUG_LOG_REQUESTS:false OLLAMA_EDITOR: OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:/var/lib/ollama OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NO_CLOUD:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false OLLAMA_VULKAN:false ROCR_VISIBLE_DEVICES: http_proxy: https_proxy: no_proxy:]"
time=2026-04-04T08:09:23.746Z level=INFO source=routes.go:1746 msg="Ollama cloud disabled: false"
time=2026-04-04T08:09:23.747Z level=INFO source=images.go:499 msg="total blobs: 29"
time=2026-04-04T08:09:23.747Z level=INFO source=images.go:506 msg="total unused blobs removed: 0"
time=2026-04-04T08:09:23.748Z level=INFO source=routes.go:1802 msg="Listening on 127.0.0.1:11434 (version 0.20.0)"
time=2026-04-04T08:09:23.748Z level=INFO source=runner.go:67 msg="discovering available GPUs..."
time=2026-04-04T08:09:23.748Z level=INFO source=server.go:432 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 44129"
time=2026-04-04T08:09:23.772Z 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="31.3 GiB" available="22.1 GiB"
time=2026-04-04T08:09:23.772Z level=INFO source=routes.go:1852 msg="vram-based default context" total_vram="0 B" default_num_ctx=4096

OS

Linux

GPU

AMD, Nvidia

CPU

AMD

Ollama version

0.20.0

Originally created by @relzp on GitHub (Apr 4, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/15321 ### What is the issue? Following the docs at https://docs.ollama.com/gpu#vulkan-gpu-support, I added the perfmon capability to the ollama executable using the command `sudo setcap cap_perfmon+ep /usr/bin/ollama` to give ollama access to the available VRAM data. However when restarting the ollama service, I noticed that neither of my GPUs were detected. I usually use Vulkan, but I also tested with ROCm, and I noticed the same behaviour (GPU detected without cap_perfmon, not detected with cap_perfmon). ### Relevant log output ```shell without cap_perfmon: Started Ollama Service. time=2026-04-04T08:08:52.131Z level=INFO source=routes.go:1744 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:0 OLLAMA_DEBUG:INFO OLLAMA_DEBUG_LOG_REQUESTS:false OLLAMA_EDITOR: OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:/var/lib/ollama OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NO_CLOUD:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false OLLAMA_VULKAN:false ROCR_VISIBLE_DEVICES: http_proxy: https_proxy: no_proxy:]" time=2026-04-04T08:08:52.131Z level=INFO source=routes.go:1746 msg="Ollama cloud disabled: false" time=2026-04-04T08:08:52.132Z level=INFO source=images.go:499 msg="total blobs: 29" time=2026-04-04T08:08:52.132Z level=INFO source=images.go:506 msg="total unused blobs removed: 0" time=2026-04-04T08:08:52.133Z level=INFO source=routes.go:1802 msg="Listening on 127.0.0.1:11434 (version 0.20.0)" time=2026-04-04T08:08:52.133Z level=INFO source=runner.go:67 msg="discovering available GPUs..." time=2026-04-04T08:08:52.133Z level=INFO source=server.go:432 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 36717" time=2026-04-04T08:08:52.268Z level=INFO source=types.go:42 msg="inference compute" id=00000000-0c00-0000-0000-000000000000 filter_id="" library=Vulkan compute=0.0 name=Vulkan0 description="AMD Radeon RX 7700 XT (RADV NAVI32)" libdirs=ollama driver=0.0 pci_id=0000:0c:00.0 type=discrete total="12.0 GiB" available="9.1 GiB" time=2026-04-04T08:08:52.268Z level=INFO source=types.go:42 msg="inference compute" id=1736597e-c3c7-5b7d-9882-ade1b4a6ea1b filter_id="" library=Vulkan compute=0.0 name=Vulkan1 description="NVIDIA GeForce GTX 1060 6GB" libdirs=ollama driver=0.0 pci_id=0000:0d:00.0 type=discrete total="6.0 GiB" available="5.5 GiB" time=2026-04-04T08:08:52.268Z level=INFO source=routes.go:1852 msg="vram-based default context" total_vram="18.0 GiB" default_num_ctx=4096 with cap_perfmon: Started Ollama Service. time=2026-04-04T08:09:23.746Z level=INFO source=routes.go:1744 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:0 OLLAMA_DEBUG:INFO OLLAMA_DEBUG_LOG_REQUESTS:false OLLAMA_EDITOR: OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:/var/lib/ollama OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NO_CLOUD:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false OLLAMA_VULKAN:false ROCR_VISIBLE_DEVICES: http_proxy: https_proxy: no_proxy:]" time=2026-04-04T08:09:23.746Z level=INFO source=routes.go:1746 msg="Ollama cloud disabled: false" time=2026-04-04T08:09:23.747Z level=INFO source=images.go:499 msg="total blobs: 29" time=2026-04-04T08:09:23.747Z level=INFO source=images.go:506 msg="total unused blobs removed: 0" time=2026-04-04T08:09:23.748Z level=INFO source=routes.go:1802 msg="Listening on 127.0.0.1:11434 (version 0.20.0)" time=2026-04-04T08:09:23.748Z level=INFO source=runner.go:67 msg="discovering available GPUs..." time=2026-04-04T08:09:23.748Z level=INFO source=server.go:432 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 44129" time=2026-04-04T08:09:23.772Z 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="31.3 GiB" available="22.1 GiB" time=2026-04-04T08:09:23.772Z level=INFO source=routes.go:1852 msg="vram-based default context" total_vram="0 B" default_num_ctx=4096 ``` ### OS Linux ### GPU AMD, Nvidia ### CPU AMD ### Ollama version 0.20.0
GiteaMirror added the bug label 2026-05-05 02:45:50 -05:00
GiteaMirror commented 2026-05-05 02:45:52 -05:00
Author
Owner
Copy Link

@be-eitel commented on GitHub (Apr 6, 2026):

OLLAMA_VULKAN:false is being logged at msg=server config, suggesting that the corresponding environment variable is not set.

Have you tried setting the OLLAMA_VULKAN env variable to 1 like noted here?

<!-- gh-comment-id:4189893996 --> @be-eitel commented on GitHub (Apr 6, 2026): `OLLAMA_VULKAN:false` is being logged at `msg=server config`, suggesting that the corresponding environment variable is not set. Have you tried setting the OLLAMA_VULKAN env variable to 1 like noted [here](https://docs.ollama.com/gpu#vulkan-gpu-support)?
GiteaMirror commented 2026-05-05 02:45:54 -05:00
Author
Owner
Copy Link

@relzp commented on GitHub (Apr 6, 2026):

Sorry yeah I do normally have it set to 1, I just happened at that time to be testing out if I needed that env to use vulkan.
I've set OLLAMA_VULKAN=1, and unfortunately it still doesn't find either GPU when cap_perfmon is set on the ollama binary.
These are the logs with OLLAMA_VULKAN=1 and OLLAMA_DEBUG=2:

Started Ollama Service.
time=2026-04-06T10:16:47.341Z level=INFO source=routes.go:1744 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:0 OLLAMA_DEBUG:DEBUG-4 OLLAMA_DEBUG_LOG_REQUESTS:false OLLAMA_EDITOR: OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:/var/lib/ollama OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NO_CLOUD:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false OLLAMA_VULKAN:true ROCR_VISIBLE_DEVICES: http_proxy: https_proxy: no_proxy:]"
time=2026-04-06T10:16:47.341Z level=INFO source=routes.go:1746 msg="Ollama cloud disabled: false"
time=2026-04-06T10:16:47.342Z level=INFO source=images.go:499 msg="total blobs: 41"
time=2026-04-06T10:16:47.343Z level=INFO source=images.go:506 msg="total unused blobs removed: 0"
time=2026-04-06T10:16:47.343Z level=INFO source=routes.go:1802 msg="Listening on 127.0.0.1:11434 (version 0.20.2)"
time=2026-04-06T10:16:47.343Z level=DEBUG source=sched.go:145 msg="starting llm scheduler"
time=2026-04-06T10:16:47.343Z level=INFO source=runner.go:67 msg="discovering available GPUs..."
time=2026-04-06T10:16:47.343Z level=TRACE source=runner.go:440 msg="starting runner for device discovery" libDirs=[/usr/lib/ollama] extraEnvs=map[]
time=2026-04-06T10:16:47.344Z level=INFO source=server.go:432 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 43729"
time=2026-04-06T10:16:47.344Z level=DEBUG source=server.go:433 msg=subprocess PATH=/usr/local/sbin:/usr/local/bin:/usr/bin OLLAMA_MODELS=/var/lib/ollama OLLAMA_VULKAN=1 OLLAMA_DEBUG=2 LD_LIBRARY_PATH=/usr/lib/ollama OLLAMA_LIBRARY_PATH=/usr/lib/ollama
time=2026-04-06T10:16:47.357Z level=INFO source=runner.go:1417 msg="starting ollama engine"
time=2026-04-06T10:16:47.358Z level=INFO source=runner.go:1452 msg="Server listening on 127.0.0.1:43729"
time=2026-04-06T10:16:47.365Z level=DEBUG source=gguf.go:604 msg=general.architecture type=string
time=2026-04-06T10:16:47.365Z level=DEBUG source=gguf.go:604 msg=tokenizer.ggml.model type=string
time=2026-04-06T10:16:47.365Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=general.alignment default=32
time=2026-04-06T10:16:47.365Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=general.alignment default=32
time=2026-04-06T10:16:47.365Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=general.file_type default=0
time=2026-04-06T10:16:47.365Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=general.name default=""
time=2026-04-06T10:16:47.365Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=general.description default=""
time=2026-04-06T10:16:47.365Z level=INFO source=ggml.go:136 msg="" architecture=llama file_type=unknown name="" description="" num_tensors=0 num_key_values=3
time=2026-04-06T10:16:47.365Z level=DEBUG source=ggml.go:94 msg="ggml backend load all from path" path=/usr/lib/ollama
time=2026-04-06T10:16:47.367Z level=INFO source=ggml.go:104 msg=system CPU.0.LLAMAFILE=1 compiler=cgo(gcc)
time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.block_count default=0
time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.pooling_type default=0
time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.expert_count default=0
time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.tokens default="&{size:0 values:[]}"
time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.scores default="&{size:0 values:[]}"
time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.token_type default="&{size:0 values:[]}"
time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.merges default="&{size:0 values:[]}"
time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.add_bos_token default=true
time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.bos_token_id default=0
time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.add_eos_token default=false
time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.eos_token_id default=0
time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.eos_token_ids default="&{size:0 values:[]}"
time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.pre default=""
time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.block_count default=0
time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.embedding_length default=0
time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.attention.head_count default=0
time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.attention.head_count_kv default=0
time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.attention.key_length default=0
time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.rope.dimension_count default=0
time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.attention.layer_norm_rms_epsilon default=0
time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.rope.freq_base default=100000
time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.rope.scaling.factor default=1
time=2026-04-06T10:16:47.368Z level=DEBUG source=runner.go:1392 msg="dummy model load took" duration=2.649655ms
time=2026-04-06T10:16:47.368Z level=DEBUG source=runner.go:1397 msg="gathering device infos took" duration=370ns
time=2026-04-06T10:16:47.368Z level=TRACE source=runner.go:467 msg="runner enumerated devices" OLLAMA_LIBRARY_PATH=[/usr/lib/ollama] devices=[]
time=2026-04-06T10:16:47.368Z level=DEBUG source=runner.go:437 msg="bootstrap discovery took" duration=24.876934ms OLLAMA_LIBRARY_PATH=[/usr/lib/ollama] extra_envs=map[]
time=2026-04-06T10:16:47.368Z level=DEBUG source=runner.go:124 msg="evaluating which, if any, devices to filter out" initial_count=0
time=2026-04-06T10:16:47.368Z level=TRACE source=runner.go:174 msg="supported GPU library combinations before filtering" supported=map[]
time=2026-04-06T10:16:47.368Z level=DEBUG source=runner.go:40 msg="GPU bootstrap discovery took" duration=25.032096ms
time=2026-04-06T10:16:47.368Z 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="31.3 GiB" available="22.8 GiB"
time=2026-04-06T10:16:47.368Z level=INFO source=routes.go:1852 msg="vram-based default context" total_vram="0 B" default_num_ctx=4096

Just to make sure:

> sudo getcap /usr/bin/ollama
/usr/bin/ollama cap_perfmon=ep
<!-- gh-comment-id:4191720196 --> @relzp commented on GitHub (Apr 6, 2026): Sorry yeah I do normally have it set to 1, I just happened at that time to be testing out if I needed that env to use vulkan. I've set OLLAMA_VULKAN=1, and unfortunately it still doesn't find either GPU when cap_perfmon is set on the ollama binary. These are the logs with OLLAMA_VULKAN=1 and OLLAMA_DEBUG=2: ``` Started Ollama Service. time=2026-04-06T10:16:47.341Z level=INFO source=routes.go:1744 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:0 OLLAMA_DEBUG:DEBUG-4 OLLAMA_DEBUG_LOG_REQUESTS:false OLLAMA_EDITOR: OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:/var/lib/ollama OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NO_CLOUD:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false OLLAMA_VULKAN:true ROCR_VISIBLE_DEVICES: http_proxy: https_proxy: no_proxy:]" time=2026-04-06T10:16:47.341Z level=INFO source=routes.go:1746 msg="Ollama cloud disabled: false" time=2026-04-06T10:16:47.342Z level=INFO source=images.go:499 msg="total blobs: 41" time=2026-04-06T10:16:47.343Z level=INFO source=images.go:506 msg="total unused blobs removed: 0" time=2026-04-06T10:16:47.343Z level=INFO source=routes.go:1802 msg="Listening on 127.0.0.1:11434 (version 0.20.2)" time=2026-04-06T10:16:47.343Z level=DEBUG source=sched.go:145 msg="starting llm scheduler" time=2026-04-06T10:16:47.343Z level=INFO source=runner.go:67 msg="discovering available GPUs..." time=2026-04-06T10:16:47.343Z level=TRACE source=runner.go:440 msg="starting runner for device discovery" libDirs=[/usr/lib/ollama] extraEnvs=map[] time=2026-04-06T10:16:47.344Z level=INFO source=server.go:432 msg="starting runner" cmd="/usr/bin/ollama runner --ollama-engine --port 43729" time=2026-04-06T10:16:47.344Z level=DEBUG source=server.go:433 msg=subprocess PATH=/usr/local/sbin:/usr/local/bin:/usr/bin OLLAMA_MODELS=/var/lib/ollama OLLAMA_VULKAN=1 OLLAMA_DEBUG=2 LD_LIBRARY_PATH=/usr/lib/ollama OLLAMA_LIBRARY_PATH=/usr/lib/ollama time=2026-04-06T10:16:47.357Z level=INFO source=runner.go:1417 msg="starting ollama engine" time=2026-04-06T10:16:47.358Z level=INFO source=runner.go:1452 msg="Server listening on 127.0.0.1:43729" time=2026-04-06T10:16:47.365Z level=DEBUG source=gguf.go:604 msg=general.architecture type=string time=2026-04-06T10:16:47.365Z level=DEBUG source=gguf.go:604 msg=tokenizer.ggml.model type=string time=2026-04-06T10:16:47.365Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=general.alignment default=32 time=2026-04-06T10:16:47.365Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=general.alignment default=32 time=2026-04-06T10:16:47.365Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=general.file_type default=0 time=2026-04-06T10:16:47.365Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=general.name default="" time=2026-04-06T10:16:47.365Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=general.description default="" time=2026-04-06T10:16:47.365Z level=INFO source=ggml.go:136 msg="" architecture=llama file_type=unknown name="" description="" num_tensors=0 num_key_values=3 time=2026-04-06T10:16:47.365Z level=DEBUG source=ggml.go:94 msg="ggml backend load all from path" path=/usr/lib/ollama time=2026-04-06T10:16:47.367Z level=INFO source=ggml.go:104 msg=system CPU.0.LLAMAFILE=1 compiler=cgo(gcc) time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.block_count default=0 time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.pooling_type default=0 time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.expert_count default=0 time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.tokens default="&{size:0 values:[]}" time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.scores default="&{size:0 values:[]}" time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.token_type default="&{size:0 values:[]}" time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.merges default="&{size:0 values:[]}" time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.add_bos_token default=true time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.bos_token_id default=0 time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.add_eos_token default=false time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.eos_token_id default=0 time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.eos_token_ids default="&{size:0 values:[]}" time=2026-04-06T10:16:47.367Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=tokenizer.ggml.pre default="" time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.block_count default=0 time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.embedding_length default=0 time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.attention.head_count default=0 time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.attention.head_count_kv default=0 time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.attention.key_length default=0 time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.rope.dimension_count default=0 time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.attention.layer_norm_rms_epsilon default=0 time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.rope.freq_base default=100000 time=2026-04-06T10:16:47.368Z level=DEBUG source=ggml.go:325 msg="key with type not found" key=llama.rope.scaling.factor default=1 time=2026-04-06T10:16:47.368Z level=DEBUG source=runner.go:1392 msg="dummy model load took" duration=2.649655ms time=2026-04-06T10:16:47.368Z level=DEBUG source=runner.go:1397 msg="gathering device infos took" duration=370ns time=2026-04-06T10:16:47.368Z level=TRACE source=runner.go:467 msg="runner enumerated devices" OLLAMA_LIBRARY_PATH=[/usr/lib/ollama] devices=[] time=2026-04-06T10:16:47.368Z level=DEBUG source=runner.go:437 msg="bootstrap discovery took" duration=24.876934ms OLLAMA_LIBRARY_PATH=[/usr/lib/ollama] extra_envs=map[] time=2026-04-06T10:16:47.368Z level=DEBUG source=runner.go:124 msg="evaluating which, if any, devices to filter out" initial_count=0 time=2026-04-06T10:16:47.368Z level=TRACE source=runner.go:174 msg="supported GPU library combinations before filtering" supported=map[] time=2026-04-06T10:16:47.368Z level=DEBUG source=runner.go:40 msg="GPU bootstrap discovery took" duration=25.032096ms time=2026-04-06T10:16:47.368Z 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="31.3 GiB" available="22.8 GiB" time=2026-04-06T10:16:47.368Z level=INFO source=routes.go:1852 msg="vram-based default context" total_vram="0 B" default_num_ctx=4096 ``` Just to make sure: ``` > sudo getcap /usr/bin/ollama /usr/bin/ollama cap_perfmon=ep ```
GiteaMirror commented 2026-05-05 02:45:54 -05:00
Author
Owner
Copy Link

@yann-amd commented on GitHub (Apr 16, 2026):

Looking at your setup, the issue is that setcap cap_perfmon+ep triggers Linux's AT_SECURE mode for the binary. When AT_SECURE is active, the dynamic linker (ld.so) strips environment variables like LD_LIBRARY_PATH for security, which is what ollama uses to find its GPU libraries at runtime.

This affects both ROCm and Vulkan backends since they both rely on library paths that get stripped. A few options that avoid this:

  • Run ollama via systemd with AmbientCapabilities=CAP_PERFMON in the service unit. systemd grants the capability at runtime without setcap, so AT_SECURE doesn't activate. Note: if you've already run setcap on the binary, you'll need sudo setcap -r /usr/bin/ollama to remove the file capabilities first, otherwise AT_SECURE still triggers.
  • Set the library search path system-wide via /etc/ld.so.conf.d/ instead of LD_LIBRARY_PATH. ldconfig paths aren't affected by AT_SECURE.
  • Use sysctl kernel.perf_event_paranoid=-1 to allow perf events without needing the capability at all (broader security tradeoff, it opens perf to all users).
<!-- gh-comment-id:4262201167 --> @yann-amd commented on GitHub (Apr 16, 2026): Looking at your setup, the issue is that `setcap cap_perfmon+ep` triggers Linux's AT_SECURE mode for the binary. When AT_SECURE is active, the dynamic linker (`ld.so`) strips environment variables like `LD_LIBRARY_PATH` for security, which is what ollama uses to find its GPU libraries at runtime. This affects both ROCm and Vulkan backends since they both rely on library paths that get stripped. A few options that avoid this: - Run ollama via systemd with `AmbientCapabilities=CAP_PERFMON` in the service unit. systemd grants the capability at runtime without `setcap`, so AT_SECURE doesn't activate. Note: if you've already run `setcap` on the binary, you'll need `sudo setcap -r /usr/bin/ollama` to remove the file capabilities first, otherwise AT_SECURE still triggers. - Set the library search path system-wide via `/etc/ld.so.conf.d/` instead of `LD_LIBRARY_PATH`. `ldconfig` paths aren't affected by AT_SECURE. - Use `sysctl kernel.perf_event_paranoid=-1` to allow perf events without needing the capability at all (broader security tradeoff, it opens perf to all users).
GiteaMirror commented 2026-05-05 02:45:56 -05:00
Author
Owner
Copy Link

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

🤖 Automated Triage & Analysis Report

Issue: #15321
Analyzed: 2026-04-18T18:22:36.971813

Analysis

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

Implementation Plan

  • Effort: medium
  • Steps:

This issue has been triaged and marked for implementation.

<!-- gh-comment-id:4274310312 --> @PureBlissAK commented on GitHub (Apr 18, 2026): <!-- ollama-issue-orchestrator:v1 issue:15321 --> ## 🤖 Automated Triage & Analysis Report **Issue**: #15321 **Analyzed**: 2026-04-18T18:22:36.971813 ### 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#71860
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?