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 #14731] 有显卡有驱动就是无法调用 #35286

New Issue
Open
opened 2026-04-22 19:40:58 -05:00 by GiteaMirror · 17 comments
GiteaMirror commented 2026-04-22 19:40:58 -05:00
Owner
Copy Link

Originally created by @kittyzero520 on GitHub (Mar 9, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/14731

What is the issue?

无法调用显卡

Relevant log output

(base) root@kittyzero:~# ollama ps
NAME              ID              SIZE      PROCESSOR    CONTEXT    UNTIL              
qwen3.5:latest    6488c96fa5fa    8.5 GB    100% CPU     4096       3 minutes from now  

(base) root@kittyzero:~# rocminfo | grep gfx
  Name:                    gfx1100                            
      Name:                    amdgcn-amd-amdhsa--gfx1100         
      Name:                    amdgcn-amd-amdhsa--gfx11-generic   
(base) root@kittyzero:~# dpkg -l|grep rocm
ii  rocm                                             7.2.0.70200-43~24.04                     amd64        Radeon Open Compute (ROCm) software stack meta package
ii  rocm-cmake                                       0.14.0.70200-43~24.04                    amd64        rocm-cmake built using CMake
ii  rocm-core                                        7.2.0.70200-43~24.04                     amd64        ROCm Runtime software stack
ii  rocm-dbgapi                                      0.77.4.70200-43~24.04                    amd64        Library to provide AMD GPU debugger API
ii  rocm-debug-agent                                 2.1.0.70200-43~24.04                     amd64        Radeon Open Compute Debug Agent (ROCdebug-agent)
ii  rocm-developer-tools                             7.2.0.70200-43~24.04                     amd64        Radeon Open Compute (ROCm) Runtime software stack
ii  rocm-device-libs                                 1.0.0.70200-43~24.04                     amd64        Radeon Open Compute - device libraries
ii  rocm-gdb                                         16.3.70200-43~24.04                      amd64        ROCgdb
ii  rocm-hip                                         7.2.0.70200-43~24.04                     amd64        Radeon Open Compute (ROCm) Runtime software stack
ii  rocm-llvm                                        22.0.0.26014.70200-43~24.04              amd64        ROCm core compiler
ii  rocm-opencl                                      2.0.0.70200-43~24.04                     amd64        clr built using CMake
ii  rocm-opencl-dev                                  2.0.0.70200-43~24.04                     amd64        clr built using CMake
ii  rocm-opencl-sdk                                  7.2.0.70200-43~24.04                     amd64        Radeon Open Compute (ROCm) Runtime software stack
ii  rocm-openmp                                      7.2.0.70200-43~24.04                     amd64        Radeon Open Compute (ROCm) OpenMP Software development Kit.
ii  rocm-smi-lib                                     7.8.0.70200-43~24.04                     amd64        AMD System Management libraries
ii  rocminfo

OS

Linux

GPU

AMD

CPU

Intel

Ollama version

v0.17.7

Originally created by @kittyzero520 on GitHub (Mar 9, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/14731 ### What is the issue? 无法调用显卡 ### Relevant log output ```shell (base) root@kittyzero:~# ollama ps NAME ID SIZE PROCESSOR CONTEXT UNTIL qwen3.5:latest 6488c96fa5fa 8.5 GB 100% CPU 4096 3 minutes from now (base) root@kittyzero:~# rocminfo | grep gfx Name: gfx1100 Name: amdgcn-amd-amdhsa--gfx1100 Name: amdgcn-amd-amdhsa--gfx11-generic (base) root@kittyzero:~# dpkg -l|grep rocm ii rocm 7.2.0.70200-43~24.04 amd64 Radeon Open Compute (ROCm) software stack meta package ii rocm-cmake 0.14.0.70200-43~24.04 amd64 rocm-cmake built using CMake ii rocm-core 7.2.0.70200-43~24.04 amd64 ROCm Runtime software stack ii rocm-dbgapi 0.77.4.70200-43~24.04 amd64 Library to provide AMD GPU debugger API ii rocm-debug-agent 2.1.0.70200-43~24.04 amd64 Radeon Open Compute Debug Agent (ROCdebug-agent) ii rocm-developer-tools 7.2.0.70200-43~24.04 amd64 Radeon Open Compute (ROCm) Runtime software stack ii rocm-device-libs 1.0.0.70200-43~24.04 amd64 Radeon Open Compute - device libraries ii rocm-gdb 16.3.70200-43~24.04 amd64 ROCgdb ii rocm-hip 7.2.0.70200-43~24.04 amd64 Radeon Open Compute (ROCm) Runtime software stack ii rocm-llvm 22.0.0.26014.70200-43~24.04 amd64 ROCm core compiler ii rocm-opencl 2.0.0.70200-43~24.04 amd64 clr built using CMake ii rocm-opencl-dev 2.0.0.70200-43~24.04 amd64 clr built using CMake ii rocm-opencl-sdk 7.2.0.70200-43~24.04 amd64 Radeon Open Compute (ROCm) Runtime software stack ii rocm-openmp 7.2.0.70200-43~24.04 amd64 Radeon Open Compute (ROCm) OpenMP Software development Kit. ii rocm-smi-lib 7.8.0.70200-43~24.04 amd64 AMD System Management libraries ii rocminfo ``` ### OS Linux ### GPU AMD ### CPU Intel ### Ollama version v0.17.7
GiteaMirror added the bug label 2026-04-22 19:40:58 -05:00
GiteaMirror commented 2026-04-22 19:40:59 -05:00
Author
Owner
Copy Link

@kittyzero520 commented on GitHub (Mar 9, 2026):

Image
<!-- gh-comment-id:4023030362 --> @kittyzero520 commented on GitHub (Mar 9, 2026): <img width="2396" height="952" alt="Image" src="https://github.com/user-attachments/assets/ab301be4-6f19-4452-be1c-48efed8f84b3" />
GiteaMirror commented 2026-04-22 19:40:59 -05:00
Author
Owner
Copy Link

@kittyzero520 commented on GitHub (Mar 9, 2026):

Image
<!-- gh-comment-id:4023036107 --> @kittyzero520 commented on GitHub (Mar 9, 2026): <img width="1796" height="100" alt="Image" src="https://github.com/user-attachments/assets/253be2d8-384a-4047-85a2-49d9a3e88ca3" />
GiteaMirror commented 2026-04-22 19:41:01 -05:00
Author
Owner
Copy Link

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

What GPU card?

<!-- gh-comment-id:4023720485 --> @rick-github commented on GitHub (Mar 9, 2026): What GPU card?
GiteaMirror commented 2026-04-22 19:41:01 -05:00
Author
Owner
Copy Link

@kittyzero520 commented on GitHub (Mar 9, 2026):

What GPU card?

amd 7900xtx 24G

(base) root@kittyzero:~# rocminfo | grep gfx
Name: gfx1100
Name: amdgcn-amd-amdhsa--gfx1100
Name: amdgcn-amd-amdhsa--gfx11-generic

<!-- gh-comment-id:4024169251 --> @kittyzero520 commented on GitHub (Mar 9, 2026): > What GPU card? amd 7900xtx 24G (base) root@kittyzero:~# rocminfo | grep gfx Name: gfx1100 Name: amdgcn-amd-amdhsa--gfx1100 Name: amdgcn-amd-amdhsa--gfx11-generic
GiteaMirror commented 2026-04-22 19:41:02 -05:00
Author
Owner
Copy Link

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

Try the Vulkan accelerator.

<!-- gh-comment-id:4024264661 --> @rick-github commented on GitHub (Mar 9, 2026): Try the [Vulkan](https://docs.ollama.com/gpu#vulkan-gpu-support) accelerator.
GiteaMirror commented 2026-04-22 19:41:04 -05:00
Author
Owner
Copy Link

@kittyzero520 commented on GitHub (Mar 9, 2026):

What GPU card?

time=2026-03-09T23:06:45.482+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 33345"
time=2026-03-09T23:06:45.549+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 44173"
time=2026-03-09T23:06:49.208+08:00 level=INFO source=runner.go:106 msg="experimental Vulkan support disabled. To enable, set OLLAMA_VULKAN=1"
time=2026-03-09T23:06:49.208+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 36395"
time=2026-03-09T23:06:49.251+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 42499"
time=2026-03-09T23:07:19.252+08:00 level=INFO source=runner.go:464 msg="failure during GPU discovery" OLLAMA_LIBRARY_PATH="[/usr/local/lib/ollama /usr/local/lib/ollama/rocm]" extra_envs="map[GGML_CUDA_INIT:1 ROCR_VISIBLE_DEVICES:GPU-6c19dc557b10e676]" error="failed to finish discovery before timeout"
time=2026-03-09T23:07:19.252+08:00 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="15.5 GiB" available="14.1 GiB"
time=2026-03-09T23:07:19.252+08:00 level=INFO source=routes.go:1763 msg="vram-based default context" total_vram="0 B" default_num_ctx=4096
[GIN] 2026/03/09 - 23:09:16 | 200 | 3.24016ms | 127.0.0.1 | HEAD "/"
[GIN] 2026/03/09 - 23:09:16 | 200 | 789.821µs | 127.0.0.1 | GET "/api/tags"
[GIN] 2026/03/09 - 23:09:26 | 200 | 28.565µs | 127.0.0.1 | HEAD "/"
[GIN] 2026/03/09 - 23:09:26 | 200 | 394.170705ms | 127.0.0.1 | POST "/api/show"
[GIN] 2026/03/09 - 23:09:27 | 200 | 437.20574ms | 127.0.0.1 | POST "/api/show"
time=2026-03-09T23:09:28.159+08:00 level=INFO source=server.go:246 msg="enabling flash attention"
time=2026-03-09T23:09:28.159+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --model /root/.ollama/models/blobs/sha256-dec52a44569a2a25341c4e4d3fee25846eed4f6f0b936278e3a3c900bb99d37c --port 46523"
time=2026-03-09T23:09:28.159+08:00 level=INFO source=sched.go:489 msg="system memory" total="15.5 GiB" free="14.1 GiB" free_swap="14.9 GiB"
time=2026-03-09T23:09:28.159+08:00 level=INFO source=server.go:757 msg="loading model" "model layers"=33 requested=-1
time=2026-03-09T23:09:28.176+08:00 level=INFO source=runner.go:1429 msg="starting ollama engine"
time=2026-03-09T23:09:28.176+08:00 level=INFO source=runner.go:1464 msg="Server listening on 127.0.0.1:46523"
time=2026-03-09T23:09:28.181+08:00 level=INFO source=runner.go:1302 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:[] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
time=2026-03-09T23:09:28.277+08:00 level=INFO source=ggml.go:136 msg="" architecture=qwen35 file_type=Q4_K_M name="" description="" num_tensors=883 num_key_values=52
load_backend: loaded CPU backend from /usr/local/lib/ollama/libggml-cpu-sandybridge.so
time=2026-03-09T23:09:28.286+08:00 level=INFO source=ggml.go:104 msg=system CPU.0.SSE3=1 CPU.0.SSSE3=1 CPU.0.AVX=1 CPU.0.LLAMAFILE=1 CPU.1.LLAMAFILE=1 compiler=cgo(gcc)
time=2026-03-09T23:09:28.980+08:00 level=INFO source=runner.go:1302 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:[] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
time=2026-03-09T23:09:30.368+08:00 level=INFO source=runner.go:1302 msg=load request="{Operation:commit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:[] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
time=2026-03-09T23:09:30.368+08:00 level=INFO source=ggml.go:482 msg="offloading 0 repeating layers to GPU"
time=2026-03-09T23:09:30.368+08:00 level=INFO source=ggml.go:486 msg="offloading output layer to CPU"
time=2026-03-09T23:09:30.368+08:00 level=INFO source=ggml.go:494 msg="offloaded 0/33 layers to GPU"
time=2026-03-09T23:09:30.368+08:00 level=INFO source=device.go:245 msg="model weights" device=CPU size="6.1 GiB"
time=2026-03-09T23:09:30.368+08:00 level=INFO source=device.go:256 msg="kv cache" device=CPU size="1.4 GiB"
time=2026-03-09T23:09:30.368+08:00 level=INFO source=device.go:267 msg="compute graph" device=CPU size="433.7 MiB"
time=2026-03-09T23:09:30.368+08:00 level=INFO source=device.go:272 msg="total memory" size="7.9 GiB"
time=2026-03-09T23:09:30.368+08:00 level=INFO source=sched.go:565 msg="loaded runners" count=1
time=2026-03-09T23:09:30.371+08:00 level=INFO source=server.go:1350 msg="waiting for llama runner to start responding"
time=2026-03-09T23:09:30.371+08:00 level=INFO source=server.go:1384 msg="waiting for server to become available" status="llm server loading model"
time=2026-03-09T23:09:46.201+08:00 level=INFO source=server.go:1388 msg="llama runner started in 18.04 seconds"
[GIN] 2026/03/09 - 23:09:46 | 200 | 18.98180016s | 127.0.0.1 | POST "/api/generate"

<!-- gh-comment-id:4024464142 --> @kittyzero520 commented on GitHub (Mar 9, 2026): > What GPU card? time=2026-03-09T23:06:45.482+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 33345" time=2026-03-09T23:06:45.549+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 44173" time=2026-03-09T23:06:49.208+08:00 level=INFO source=runner.go:106 msg="experimental Vulkan support disabled. To enable, set OLLAMA_VULKAN=1" time=2026-03-09T23:06:49.208+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 36395" time=2026-03-09T23:06:49.251+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 42499" time=2026-03-09T23:07:19.252+08:00 level=INFO source=runner.go:464 msg="failure during GPU discovery" OLLAMA_LIBRARY_PATH="[/usr/local/lib/ollama /usr/local/lib/ollama/rocm]" extra_envs="map[GGML_CUDA_INIT:1 ROCR_VISIBLE_DEVICES:GPU-6c19dc557b10e676]" error="failed to finish discovery before timeout" time=2026-03-09T23:07:19.252+08:00 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="15.5 GiB" available="14.1 GiB" time=2026-03-09T23:07:19.252+08:00 level=INFO source=routes.go:1763 msg="vram-based default context" total_vram="0 B" default_num_ctx=4096 [GIN] 2026/03/09 - 23:09:16 | 200 | 3.24016ms | 127.0.0.1 | HEAD "/" [GIN] 2026/03/09 - 23:09:16 | 200 | 789.821µs | 127.0.0.1 | GET "/api/tags" [GIN] 2026/03/09 - 23:09:26 | 200 | 28.565µs | 127.0.0.1 | HEAD "/" [GIN] 2026/03/09 - 23:09:26 | 200 | 394.170705ms | 127.0.0.1 | POST "/api/show" [GIN] 2026/03/09 - 23:09:27 | 200 | 437.20574ms | 127.0.0.1 | POST "/api/show" time=2026-03-09T23:09:28.159+08:00 level=INFO source=server.go:246 msg="enabling flash attention" time=2026-03-09T23:09:28.159+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --model /root/.ollama/models/blobs/sha256-dec52a44569a2a25341c4e4d3fee25846eed4f6f0b936278e3a3c900bb99d37c --port 46523" time=2026-03-09T23:09:28.159+08:00 level=INFO source=sched.go:489 msg="system memory" total="15.5 GiB" free="14.1 GiB" free_swap="14.9 GiB" time=2026-03-09T23:09:28.159+08:00 level=INFO source=server.go:757 msg="loading model" "model layers"=33 requested=-1 time=2026-03-09T23:09:28.176+08:00 level=INFO source=runner.go:1429 msg="starting ollama engine" time=2026-03-09T23:09:28.176+08:00 level=INFO source=runner.go:1464 msg="Server listening on 127.0.0.1:46523" time=2026-03-09T23:09:28.181+08:00 level=INFO source=runner.go:1302 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:[] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" time=2026-03-09T23:09:28.277+08:00 level=INFO source=ggml.go:136 msg="" architecture=qwen35 file_type=Q4_K_M name="" description="" num_tensors=883 num_key_values=52 load_backend: loaded CPU backend from /usr/local/lib/ollama/libggml-cpu-sandybridge.so time=2026-03-09T23:09:28.286+08:00 level=INFO source=ggml.go:104 msg=system CPU.0.SSE3=1 CPU.0.SSSE3=1 CPU.0.AVX=1 CPU.0.LLAMAFILE=1 CPU.1.LLAMAFILE=1 compiler=cgo(gcc) time=2026-03-09T23:09:28.980+08:00 level=INFO source=runner.go:1302 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:[] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" time=2026-03-09T23:09:30.368+08:00 level=INFO source=runner.go:1302 msg=load request="{Operation:commit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:[] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" time=2026-03-09T23:09:30.368+08:00 level=INFO source=ggml.go:482 msg="offloading 0 repeating layers to GPU" time=2026-03-09T23:09:30.368+08:00 level=INFO source=ggml.go:486 msg="offloading output layer to CPU" time=2026-03-09T23:09:30.368+08:00 level=INFO source=ggml.go:494 msg="offloaded 0/33 layers to GPU" time=2026-03-09T23:09:30.368+08:00 level=INFO source=device.go:245 msg="model weights" device=CPU size="6.1 GiB" time=2026-03-09T23:09:30.368+08:00 level=INFO source=device.go:256 msg="kv cache" device=CPU size="1.4 GiB" time=2026-03-09T23:09:30.368+08:00 level=INFO source=device.go:267 msg="compute graph" device=CPU size="433.7 MiB" time=2026-03-09T23:09:30.368+08:00 level=INFO source=device.go:272 msg="total memory" size="7.9 GiB" time=2026-03-09T23:09:30.368+08:00 level=INFO source=sched.go:565 msg="loaded runners" count=1 time=2026-03-09T23:09:30.371+08:00 level=INFO source=server.go:1350 msg="waiting for llama runner to start responding" time=2026-03-09T23:09:30.371+08:00 level=INFO source=server.go:1384 msg="waiting for server to become available" status="llm server loading model" time=2026-03-09T23:09:46.201+08:00 level=INFO source=server.go:1388 msg="llama runner started in 18.04 seconds" [GIN] 2026/03/09 - 23:09:46 | 200 | 18.98180016s | 127.0.0.1 | POST "/api/generate"
GiteaMirror commented 2026-04-22 19:41:06 -05:00
Author
Owner
Copy Link

@kittyzero520 commented on GitHub (Mar 10, 2026):

Try the Vulkan accelerator.

The problem is that I’ve never opened this before, and it was normal before. Why isn’t it working all of a sudden?

<!-- gh-comment-id:4027942850 --> @kittyzero520 commented on GitHub (Mar 10, 2026): > Try the [Vulkan](https://docs.ollama.com/gpu#vulkan-gpu-support) accelerator. The problem is that I’ve never opened this before, and it was normal before. Why isn’t it working all of a sudden?
GiteaMirror commented 2026-04-22 19:41:07 -05:00
Author
Owner
Copy Link

@kittyzero520 commented on GitHub (Mar 11, 2026):

(base) root@kittyzero:# export LD_LIBRARY_PATH=/opt/rocm/lib:/opt/rocm-7.2.0/lib:$LD_LIBRARY_PATH
(base) root@kittyzero:# export HSA_OVERRIDE_GFX_VERSION=11.0.0
(base) root@kittyzero:# export HIP_VISIBLE_DEVICES=0
(base) root@kittyzero:# pkill -9 ollama
(base) root@kittyzero:# ollama serve 2>&1 &
[1] 25047
(base) root@kittyzero:# sleep 5time=2026-03-11T13:14:02.047+08:00 level=INFO source=routes.go:1658 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES:0 HSA_OVERRIDE_GFX_VERSION:11.0.0 HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:0 OLLAMA_DEBUG:INFO 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:/root/.ollama/models 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=2026-03-11T13:14:02.047+08:00 level=INFO source=routes.go:1660 msg="Ollama cloud disabled: false"
time=2026-03-11T13:14:02.048+08:00 level=INFO source=images.go:477 msg="total blobs: 7"
time=2026-03-11T13:14:02.048+08:00 level=INFO source=images.go:484 msg="total unused blobs removed: 0"
time=2026-03-11T13:14:02.049+08:00 level=INFO source=routes.go:1713 msg="Listening on 127.0.0.1:11434 (version 0.17.7)"
time=2026-03-11T13:14:02.049+08:00 level=INFO source=runner.go:67 msg="discovering available GPUs..."
time=2026-03-11T13:14:02.049+08:00 level=WARN source=runner.go:485 msg="user overrode visible devices" HIP_VISIBLE_DEVICES=0
time=2026-03-11T13:14:02.049+08:00 level=WARN source=runner.go:485 msg="user overrode visible devices" HSA_OVERRIDE_GFX_VERSION=11.0.0
time=2026-03-11T13:14:02.049+08:00 level=WARN source=runner.go:489 msg="if GPUs are not correctly discovered, unset and try again"
time=2026-03-11T13:14:02.050+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 35559"
time=2026-03-11T13:14:02.081+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 45679"
time=2026-03-11T13:14:02.112+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 34845"
time=2026-03-11T13:14:04.348+08:00 level=INFO source=runner.go:106 msg="experimental Vulkan support disabled. To enable, set OLLAMA_VULKAN=1"
time=2026-03-11T13:14:04.348+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 33835"

<!-- gh-comment-id:4036501970 --> @kittyzero520 commented on GitHub (Mar 11, 2026): (base) root@kittyzero:~# export LD_LIBRARY_PATH=/opt/rocm/lib:/opt/rocm-7.2.0/lib:$LD_LIBRARY_PATH (base) root@kittyzero:~# export HSA_OVERRIDE_GFX_VERSION=11.0.0 (base) root@kittyzero:~# export HIP_VISIBLE_DEVICES=0 (base) root@kittyzero:~# pkill -9 ollama (base) root@kittyzero:~# ollama serve 2>&1 & [1] 25047 (base) root@kittyzero:~# sleep 5time=2026-03-11T13:14:02.047+08:00 level=INFO source=routes.go:1658 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES:0 HSA_OVERRIDE_GFX_VERSION:11.0.0 HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:0 OLLAMA_DEBUG:INFO 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:/root/.ollama/models 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=2026-03-11T13:14:02.047+08:00 level=INFO source=routes.go:1660 msg="Ollama cloud disabled: false" time=2026-03-11T13:14:02.048+08:00 level=INFO source=images.go:477 msg="total blobs: 7" time=2026-03-11T13:14:02.048+08:00 level=INFO source=images.go:484 msg="total unused blobs removed: 0" time=2026-03-11T13:14:02.049+08:00 level=INFO source=routes.go:1713 msg="Listening on 127.0.0.1:11434 (version 0.17.7)" time=2026-03-11T13:14:02.049+08:00 level=INFO source=runner.go:67 msg="discovering available GPUs..." time=2026-03-11T13:14:02.049+08:00 level=WARN source=runner.go:485 msg="user overrode visible devices" HIP_VISIBLE_DEVICES=0 time=2026-03-11T13:14:02.049+08:00 level=WARN source=runner.go:485 msg="user overrode visible devices" HSA_OVERRIDE_GFX_VERSION=11.0.0 time=2026-03-11T13:14:02.049+08:00 level=WARN source=runner.go:489 msg="if GPUs are not correctly discovered, unset and try again" time=2026-03-11T13:14:02.050+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 35559" time=2026-03-11T13:14:02.081+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 45679" time=2026-03-11T13:14:02.112+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 34845" time=2026-03-11T13:14:04.348+08:00 level=INFO source=runner.go:106 msg="experimental Vulkan support disabled. To enable, set OLLAMA_VULKAN=1" time=2026-03-11T13:14:04.348+08:00 level=INFO source=server.go:430 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 33835"
GiteaMirror commented 2026-04-22 19:41:08 -05:00
Author
Owner
Copy Link

@Jasdfgh commented on GitHub (Mar 12, 2026):

7900 XTX (gfx1100) 在 ollama的Linux 支持列表(https://docs.ollama.com/gpu) 里是原生支持的（LLVM target gfx1100 + Radeon RX 那边写了），GPU 发现超时不太正常。

你这几个环境变量不太对：

  • HSA_OVERRIDE_GFX_VERSION的话，是给不在支持列表里的 GPU 用的，让它伪装成一个接近的架构。gfx1100 本身就在列表里，不需要 override。设了可能导致匹配出错？
  • HIP_VISIBLE_DEVICES的话，你只有一个 GPU，好像不需要设。而且，ollama 的文档（https://docs.ollama.com/gpu）里， 好像AMD GPU 选择用的是 ROCR_VISIBLE_DEVICES，不是 HIP_VISIBLE_DEVICES。
  • 而且，ollama 日志也在警告这些 override 可能干扰检测。

我建议先把这些 env vars 全部 unset，重新启动试试。另外 0.17.8 RC 刚出，更新了 Linux ROCm v7 支持，安装方法在这：https://github.com/ollama/ollama/blob/main/docs/linux.mdx#installing-specific-versions。

还有问题可以说下是什么时候开始不行的，中间有没有更新过驱动或者内核。

<!-- gh-comment-id:4045360524 --> @Jasdfgh commented on GitHub (Mar 12, 2026): 7900 XTX (gfx1100) 在 ollama的Linux 支持列表(https://docs.ollama.com/gpu) 里是原生支持的（LLVM target gfx1100 + Radeon RX 那边写了），GPU 发现超时不太正常。 你这几个环境变量不太对： - HSA_OVERRIDE_GFX_VERSION的话，是给不在支持列表里的 GPU 用的，让它伪装成一个接近的架构。gfx1100 本身就在列表里，不需要 override。设了可能导致匹配出错？ - HIP_VISIBLE_DEVICES的话，你只有一个 GPU，好像不需要设。而且，ollama 的文档（https://docs.ollama.com/gpu）里， 好像AMD GPU 选择用的是 ROCR_VISIBLE_DEVICES，不是 HIP_VISIBLE_DEVICES。 - 而且，ollama 日志也在警告这些 override 可能干扰检测。 我建议先把这些 env vars 全部 unset，重新启动试试。另外 0.17.8 RC 刚出，更新了 Linux ROCm v7 支持，安装方法在这：https://github.com/ollama/ollama/blob/main/docs/linux.mdx#installing-specific-versions。 还有问题可以说下是什么时候开始不行的，中间有没有更新过驱动或者内核。
GiteaMirror commented 2026-04-22 19:41:09 -05:00
Author
Owner
Copy Link

@kittyzero520 commented on GitHub (Mar 13, 2026):

7900 XTX (gfx1100) 在 ollama的Linux 支持列表(https://docs.ollama.com/gpu) 里是原生支持的（LLVM target gfx1100 + Radeon RX 那边写了），GPU 发现超时不太正常。

你这几个环境变量不太对：

  • HSA_OVERRIDE_GFX_VERSION的话，是给不在支持列表里的 GPU 用的，让它伪装成一个接近的架构。gfx1100 本身就在列表里，不需要 override。设了可能导致匹配出错？
  • HIP_VISIBLE_DEVICES的话，你只有一个 GPU，好像不需要设。而且，ollama 的文档（https://docs.ollama.com/gpu）里， 好像AMD GPU 选择用的是 ROCR_VISIBLE_DEVICES，不是 HIP_VISIBLE_DEVICES。
  • 而且，ollama 日志也在警告这些 override 可能干扰检测。

我建议先把这些 env vars 全部 unset，重新启动试试。另外 0.17.8 RC 刚出，更新了 Linux ROCm v7 支持，安装方法在这：https://github.com/ollama/ollama/blob/main/docs/linux.mdx#installing-specific-versions。

还有问题可以说下是什么时候开始不行的，中间有没有更新过驱动或者内核。

发现一个问题，如果把增加下面参数，Ollama可以使用显卡；你的这个方案，我试下，感谢你的支撑！

Image

7900 XTX (gfx1100) 在 ollama的Linux 支持列表(https://docs.ollama.com/gpu) 里是原生支持的（LLVM target gfx1100 + Radeon RX 那边写了），GPU 发现超时不太正常。

你这几个环境变量不太对：

  • HSA_OVERRIDE_GFX_VERSION的话，是给不在支持列表里的 GPU 用的，让它伪装成一个接近的架构。gfx1100 本身就在列表里，不需要 override。设了可能导致匹配出错？
  • HIP_VISIBLE_DEVICES的话，你只有一个 GPU，好像不需要设。而且，ollama 的文档（https://docs.ollama.com/gpu）里， 好像AMD GPU 选择用的是 ROCR_VISIBLE_DEVICES，不是 HIP_VISIBLE_DEVICES。
  • 而且，ollama 日志也在警告这些 override 可能干扰检测。

我建议先把这些 env vars 全部 unset，重新启动试试。另外 0.17.8 RC 刚出，更新了 Linux ROCm v7 支持，安装方法在这：https://github.com/ollama/ollama/blob/main/docs/linux.mdx#installing-specific-versions。

还有问题可以说下是什么时候开始不行的，中间有没有更新过驱动或者内核。

升级到最新的17就不行了

<!-- gh-comment-id:4055606958 --> @kittyzero520 commented on GitHub (Mar 13, 2026): > 7900 XTX (gfx1100) 在 ollama的Linux 支持列表(https://docs.ollama.com/gpu) 里是原生支持的（LLVM target gfx1100 + Radeon RX 那边写了），GPU 发现超时不太正常。 > > 你这几个环境变量不太对： > > * HSA_OVERRIDE_GFX_VERSION的话，是给不在支持列表里的 GPU 用的，让它伪装成一个接近的架构。gfx1100 本身就在列表里，不需要 override。设了可能导致匹配出错？ > * HIP_VISIBLE_DEVICES的话，你只有一个 GPU，好像不需要设。而且，ollama 的文档（[https://docs.ollama.com/gpu）里，](https://docs.ollama.com/gpu%EF%BC%89%E9%87%8C%EF%BC%8C) 好像AMD GPU 选择用的是 ROCR_VISIBLE_DEVICES，不是 HIP_VISIBLE_DEVICES。 > * 而且，ollama 日志也在警告这些 override 可能干扰检测。 > > 我建议先把这些 env vars 全部 unset，重新启动试试。另外 0.17.8 RC 刚出，更新了 Linux ROCm v7 支持，安装方法在这：[https://github.com/ollama/ollama/blob/main/docs/linux.mdx#installing-specific-versions。](https://github.com/ollama/ollama/blob/main/docs/linux.mdx#installing-specific-versions%E3%80%82) > > 还有问题可以说下是什么时候开始不行的，中间有没有更新过驱动或者内核。 发现一个问题，如果把增加下面参数，Ollama可以使用显卡；你的这个方案，我试下，感谢你的支撑！ <img width="1196" height="956" alt="Image" src="https://github.com/user-attachments/assets/926ce6bb-b2a7-4499-af6e-0c041f8dd9fa" /> > 7900 XTX (gfx1100) 在 ollama的Linux 支持列表(https://docs.ollama.com/gpu) 里是原生支持的（LLVM target gfx1100 + Radeon RX 那边写了），GPU 发现超时不太正常。 > > 你这几个环境变量不太对： > > * HSA_OVERRIDE_GFX_VERSION的话，是给不在支持列表里的 GPU 用的，让它伪装成一个接近的架构。gfx1100 本身就在列表里，不需要 override。设了可能导致匹配出错？ > * HIP_VISIBLE_DEVICES的话，你只有一个 GPU，好像不需要设。而且，ollama 的文档（[https://docs.ollama.com/gpu）里，](https://docs.ollama.com/gpu%EF%BC%89%E9%87%8C%EF%BC%8C) 好像AMD GPU 选择用的是 ROCR_VISIBLE_DEVICES，不是 HIP_VISIBLE_DEVICES。 > * 而且，ollama 日志也在警告这些 override 可能干扰检测。 > > 我建议先把这些 env vars 全部 unset，重新启动试试。另外 0.17.8 RC 刚出，更新了 Linux ROCm v7 支持，安装方法在这：[https://github.com/ollama/ollama/blob/main/docs/linux.mdx#installing-specific-versions。](https://github.com/ollama/ollama/blob/main/docs/linux.mdx#installing-specific-versions%E3%80%82) > > 还有问题可以说下是什么时候开始不行的，中间有没有更新过驱动或者内核。 升级到最新的17就不行了
GiteaMirror commented 2026-04-22 19:41:10 -05:00
Author
Owner
Copy Link

@Jasdfgh commented on GitHub (Mar 15, 2026):

看到了，你那两个命令，是切到 Vulkan 绕过去了。这能用但走的不是 ROCm 路径。ROCm 那边 GPU discovery 超时的问题还在。你之前 ROCm 能正常用的时候是哪个版本的 ollama？看看是不是 0.17 引入的问题。

<!-- gh-comment-id:4063217936 --> @Jasdfgh commented on GitHub (Mar 15, 2026): 看到了，你那两个命令，是切到 Vulkan 绕过去了。这能用但走的不是 ROCm 路径。ROCm 那边 GPU discovery 超时的问题还在。你之前 ROCm 能正常用的时候是哪个版本的 ollama？看看是不是 0.17 引入的问题。
GiteaMirror commented 2026-04-22 19:41:11 -05:00
Author
Owner
Copy Link

@kittyzero520 commented on GitHub (Mar 16, 2026):

看到了，你那两个命令，是切到 Vulkan 绕过去了。这能用但走的不是 ROCm 路径。ROCm 那边 GPU discovery 超时的问题还在。你之前 ROCm 能正常用的时候是哪个版本的 ollama？看看是不是 0.17 引入的问题。

1.之前版本不记得了
2.审计0.18版本可以加载qwen3.5 9b模型了，但是27B直接让ssh丢失连接，服务器无法访问了

Image Image
<!-- gh-comment-id:4065176842 --> @kittyzero520 commented on GitHub (Mar 16, 2026): > 看到了，你那两个命令，是切到 Vulkan 绕过去了。这能用但走的不是 ROCm 路径。ROCm 那边 GPU discovery 超时的问题还在。你之前 ROCm 能正常用的时候是哪个版本的 ollama？看看是不是 0.17 引入的问题。 1.之前版本不记得了 2.审计0.18版本可以加载qwen3.5 9b模型了，但是27B直接让ssh丢失连接，服务器无法访问了 <img width="2400" height="1560" alt="Image" src="https://github.com/user-attachments/assets/ea08432d-e024-422c-8804-e0897018e02e" /> <img width="2400" height="1560" alt="Image" src="https://github.com/user-attachments/assets/73fee515-b800-40e4-9f56-3c56cc5fd576" />
GiteaMirror commented 2026-04-22 19:41:12 -05:00
Author
Owner
Copy Link

@kittyzero520 commented on GitHub (Mar 16, 2026):

看到了，你那两个命令，是切到 Vulkan 绕过去了。这能用但走的不是 ROCm 路径。ROCm 那边 GPU discovery 超时的问题还在。你之前 ROCm 能正常用的时候是哪个版本的 ollama？看看是不是 0.17 引入的问题。

1.之前版本不记得了 2.审计0.18版本可以加载qwen3.5 9b模型了，但是27B直接让ssh丢失连接，服务器无法访问了，我看ram 内存16G 突然吃满了

Image Image

<!-- gh-comment-id:4065180860 --> @kittyzero520 commented on GitHub (Mar 16, 2026): > > 看到了，你那两个命令，是切到 Vulkan 绕过去了。这能用但走的不是 ROCm 路径。ROCm 那边 GPU discovery 超时的问题还在。你之前 ROCm 能正常用的时候是哪个版本的 ollama？看看是不是 0.17 引入的问题。 > > 1.之前版本不记得了 2.审计0.18版本可以加载qwen3.5 9b模型了，但是27B直接让ssh丢失连接，服务器无法访问了，我看ram 内存16G 突然吃满了 > > <img alt="Image" width="2000" height="1560" src="https://private-user-images.githubusercontent.com/33313886/563933945-ea08432d-e024-422c-8804-e0897018e02e.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3NzM2Mzk4OTIsIm5iZiI6MTc3MzYzOTU5MiwicGF0aCI6Ii8zMzMxMzg4Ni81NjM5MzM5NDUtZWEwODQzMmQtZTAyNC00MjJjLTg4MDQtZTA4OTcwMThlMDJlLnBuZz9YLUFtei1BbGdvcml0aG09QVdTNC1ITUFDLVNIQTI1NiZYLUFtei1DcmVkZW50aWFsPUFLSUFWQ09EWUxTQTUzUFFLNFpBJTJGMjAyNjAzMTYlMkZ1cy1lYXN0LTElMkZzMyUyRmF3czRfcmVxdWVzdCZYLUFtei1EYXRlPTIwMjYwMzE2VDA1Mzk1MlomWC1BbXotRXhwaXJlcz0zMDAmWC1BbXotU2lnbmF0dXJlPWE2MTg0M2FhM2NmNTNlZTFmNDNkYzY5MjM3ZjhmNDJmM2I3NjFhZTRmMjczNTJkMjc2MDExOTBlZTVlZjBiMTYmWC1BbXotU2lnbmVkSGVhZGVycz1ob3N0In0.Bsr23D7TFPQMovcJpe56d5QXvtZGt78U9JbiYzeogmc"> <img alt="Image" width="2000" height="1560" src="https://private-user-images.githubusercontent.com/33313886/563934004-73fee515-b800-40e4-9f56-3c56cc5fd576.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3NzM2Mzk4OTIsIm5iZiI6MTc3MzYzOTU5MiwicGF0aCI6Ii8zMzMxMzg4Ni81NjM5MzQwMDQtNzNmZWU1MTUtYjgwMC00MGU0LTlmNTYtM2M1NmNjNWZkNTc2LnBuZz9YLUFtei1BbGdvcml0aG09QVdTNC1ITUFDLVNIQTI1NiZYLUFtei1DcmVkZW50aWFsPUFLSUFWQ09EWUxTQTUzUFFLNFpBJTJGMjAyNjAzMTYlMkZ1cy1lYXN0LTElMkZzMyUyRmF3czRfcmVxdWVzdCZYLUFtei1EYXRlPTIwMjYwMzE2VDA1Mzk1MlomWC1BbXotRXhwaXJlcz0zMDAmWC1BbXotU2lnbmF0dXJlPWZkN2VhNWZiNzFlMWY5MGY4ZGM3ZTZlOGQ1NTQzZTc3YTkwYjY1MmU0MGZkODEzOGIxMTg0NTIwYTEwZWZmZDMmWC1BbXotU2lnbmVkSGVhZGVycz1ob3N0In0.h8l7wd7lUBY_Z4rV7JgytwYnRs85wSJgnTe01Xl7mRU">
GiteaMirror commented 2026-04-22 19:41:13 -05:00
Author
Owner
Copy Link

@kittyzero520 commented on GitHub (Mar 16, 2026):

看到了，你那两个命令，是切到 Vulkan 绕过去了。这能用但走的不是 ROCm 路径。ROCm 那边 GPU discovery 超时的问题还在。你之前 ROCm 能正常用的时候是哪个版本的 ollama？看看是不是 0.17 引入的问题。

1.之前版本不记得了
2.现在0.18版本可以加载qwen3.5 9b模型了，但是27B直接让ssh丢失连接，服务器无法访问了，而且ram 内存16G 突然吃满了

Image Image

<!-- gh-comment-id:4065182601 --> @kittyzero520 commented on GitHub (Mar 16, 2026): > > > 看到了，你那两个命令，是切到 Vulkan 绕过去了。这能用但走的不是 ROCm 路径。ROCm 那边 GPU discovery 超时的问题还在。你之前 ROCm 能正常用的时候是哪个版本的 ollama？看看是不是 0.17 引入的问题。 1.之前版本不记得了 2.现在0.18版本可以加载qwen3.5 9b模型了，但是27B直接让ssh丢失连接，服务器无法访问了，而且ram 内存16G 突然吃满了 > > <img alt="Image" width="2000" height="1560" src="https://private-user-images.githubusercontent.com/33313886/563933945-ea08432d-e024-422c-8804-e0897018e02e.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.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.Bsr23D7TFPQMovcJpe56d5QXvtZGt78U9JbiYzeogmc"> <img alt="Image" width="2000" height="1560" src="https://private-user-images.githubusercontent.com/33313886/563934004-73fee515-b800-40e4-9f56-3c56cc5fd576.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.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.h8l7wd7lUBY_Z4rV7JgytwYnRs85wSJgnTe01Xl7mRU">
GiteaMirror commented 2026-04-22 19:41:14 -05:00
Author
Owner
Copy Link

@Jasdfgh commented on GitHub (Mar 18, 2026):

从 ollama ps 看 GPU 是在用的（100% GPU），9B 和 8B 都正常。27B 的问题，我觉得应该是 VRAM 不够了，模型本身 24GB 加上 32768 context 的 KV cache 超过了你 24G 显存，溢出到系统内存，你系统 RAM 只有 15.5G 所以直接 OOM 了。

试一下降 context，比如 OLLAMA_CONTEXT_LENGTH=4096 ollama run qwen3.5:27b，或者用小一点的量化版本？

<!-- gh-comment-id:4079875740 --> @Jasdfgh commented on GitHub (Mar 18, 2026): 从 ollama ps 看 GPU 是在用的（100% GPU），9B 和 8B 都正常。27B 的问题，我觉得应该是 VRAM 不够了，模型本身 24GB 加上 32768 context 的 KV cache 超过了你 24G 显存，溢出到系统内存，你系统 RAM 只有 15.5G 所以直接 OOM 了。 试一下降 context，比如 OLLAMA_CONTEXT_LENGTH=4096 ollama run qwen3.5:27b，或者用小一点的量化版本？
GiteaMirror commented 2026-04-22 19:41:15 -05:00
Author
Owner
Copy Link

@kittyzero520 commented on GitHub (Mar 18, 2026):

从 ollama ps 看 GPU 是在用的（100% GPU），9B 和 8B 都正常。27B 的问题，我觉得应该是 VRAM 不够了，模型本身 24GB 加上 32768 context 的 KV cache 超过了你 24G 显存，溢出到系统内存，你系统 RAM 只有 15.5G 所以直接 OOM 了。

试一下降 context，比如 OLLAMA_CONTEXT_LENGTH=4096 ollama run qwen3.5:27b，或者用小一点的量化版本？

我试一下

<!-- gh-comment-id:4082871894 --> @kittyzero520 commented on GitHub (Mar 18, 2026): > 从 ollama ps 看 GPU 是在用的（100% GPU），9B 和 8B 都正常。27B 的问题，我觉得应该是 VRAM 不够了，模型本身 24GB 加上 32768 context 的 KV cache 超过了你 24G 显存，溢出到系统内存，你系统 RAM 只有 15.5G 所以直接 OOM 了。 > > 试一下降 context，比如 OLLAMA_CONTEXT_LENGTH=4096 ollama run qwen3.5:27b，或者用小一点的量化版本？ 我试一下
GiteaMirror commented 2026-04-22 19:41:17 -05:00
Author
Owner
Copy Link

@Jasdfgh commented on GitHub (Mar 30, 2026):

如何了？后来降了 context size 或者换了小量化版本之后有好转吗？

<!-- gh-comment-id:4151823089 --> @Jasdfgh commented on GitHub (Mar 30, 2026): 如何了？后来降了 context size 或者换了小量化版本之后有好转吗？
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#35286
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?