ollama - 💡(How to fix) Fix Vulkan backend crashes with C++ exception on Intel ARC A750 when using gemma4 26b MoE + image input [1 participants]

ollama2026-04-30 03:51:13

ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

GitHub issue URL

Helpful · Quick feedback

GitHub stats

ollama/ollama#15890•Fetched 2026-05-01 05:33:33

View on GitHub

Comments

Participants

Timeline

Reactions

Author

ZhouYvChongShan

Participants

ZhouYvChongShan

Timeline (top)

labeled ×1subscribed ×1unsubscribed ×1

Error Message

0ac github.com/ollama/ollama/runner/ollamarunner.Execute.gowrap1() github.com/ollama/ollama/runner/ollamarunner/runner.go:1430 +0x28 fp=0xc00086ffe0 sp=0xc00086ffb8 pc=0x7ff686ba88e8 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc00086ffe8 sp=0xc00086ffe0 pc=0x7ff68650db21 created by github.com/ollama/ollama/runner/ollamarunner.Execute in goroutine 1 github.com/ollama/ollama/runner/ollamarunner/runner.go:1430 +0x4c9

goroutine 15 gp=0xc0004168c0 m=nil [select]: runtime.gopark(0xc000047a08?, 0x2?, 0xc0?, 0x77?, 0xc00004786c?) runtime/proc.go:435 +0xce fp=0xc000047698 sp=0xc000047678 pc=0x7ff68650598e runtime.selectgo(0xc000047a08, 0xc000047868, 0xe3c?, 0x0, 0x1?, 0x1) runtime/select.go:351 +0x837 fp=0xc0000477d0 sp=0xc000047698 pc=0x7ff6864e6437 github.com/ollama/ollama/runner/ollamarunner.(*Server).completion(0xc00051c000, {0x7ff687fa3020, 0xc000692000}, 0xc000616000) github.com/ollama/ollama/runner/ollamarunner/runner.go:956 +0xc4e fp=0xc000047ac0 sp=0xc0000477d0 pc=0x7ff686ba388e github.com/ollama/ollama/runner/ollamarunner.(*Server).completion-fm({0x7ff687fa3020?, 0xc000692000?}, 0xc000047b40?) <autogenerated>:1 +0x36 fp=0xc000047af0 sp=0xc000047ac0 pc=0x7ff686ba8dd6 net/http.HandlerFunc.ServeHTTP(0xc000520000?, {0x7ff687fa3020?, 0xc000692000?}, 0xc000047b60?) net/http/server.go:2294 +0x29 fp=0xc000047b18 sp=0xc000047af0 pc=0x7ff686825d49 net/http.(*ServeMux).ServeHTTP(0x7ff6864ab785?, {0x7ff687fa3020, 0xc000692000}, 0xc000616000) net/http/server.go:2822 +0x1c4 fp=0xc000047b68 sp=0xc000047b18 pc=0x7ff686827c44 net/http.serverHandler.ServeHTTP({0x7ff687f9eeb0?}, {0x7ff687fa3020?, 0xc000692000?}, 0x1?) net/http/server.go:3301 +0x8e fp=0xc000047b98 sp=0xc000047b68 pc=0x7ff6868456ce net/http.(*conn).serve(0xc000532090, {0x7ff687fa5978, 0xc000512390}) net/http/server.go:2102 +0x625 fp=0xc000047fb8 sp=0xc000047b98 pc=0x7ff686824245 net/http.(*Server).Serve.gowrap3() net/http/server.go:3454 +0x28 fp=0xc000047fe0 sp=0xc000047fb8 pc=0x7ff686829b08 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc000047fe8 sp=0xc000047fe0 pc=0x7ff68650db21 created by net/http.(*Server).Serve in goroutine 1 net/http/server.go:3454 +0x485

goroutine 895 gp=0xc000507dc0 m=nil [IO wait]: runtime.gopark(0x0?, 0xc00051e2a0?, 0x48?, 0xe3?, 0xc00051e34c?) runtime/proc.go:435 +0xce fp=0xc001725d58 sp=0xc001725d38 pc=0x7ff68650598e runtime.netpollblock(0x2ac?, 0x864a0406?, 0xf6?) runtime/netpoll.go:575 +0xf7 fp=0xc001725d90 sp=0xc001725d58 pc=0x7ff6864cbdf7 internal/poll.runtime_pollWait(0x1dff7d11858, 0x72) runtime/netpoll.go:351 +0x85 fp=0xc001725db0 sp=0xc001725d90 pc=0x7ff686504b25 internal/poll.(*pollDesc).wait(0x0?, 0xc000532100?, 0x0) internal/poll/fd_poll_runtime.go:84 +0x27 fp=0xc001725dd8 sp=0xc001725db0 pc=0x7ff68659c0c7 internal/poll.execIO(0xc00051e2a0, 0x7ff687dd8778) internal/poll/fd_windows.go:177 +0x105 fp=0xc001725e50 sp=0xc001725dd8 pc=0x7ff68659d525 internal/poll.(*FD).Read(0xc00051e288, {0xc0003d20a1, 0x1, 0x1}) internal/poll/fd_windows.go:438 +0x29b fp=0xc001725ef0 sp=0xc001725e50 pc=0x7ff68659e1fb net.(*netFD).Read(0xc00051e288, {0xc0003d20a1?, 0x0?, 0x0?}) net/fd_posix.go:55 +0x25 fp=0xc001725f38 sp=0xc001725ef0 pc=0x7ff686611945 net.(*conn).Read(0xc000602000, {0xc0003d20a1?, 0x0?, 0x0?}) net/net.go:194 +0x45 fp=0xc001725f80 sp=0xc001725f38 pc=0x7ff686621065 net/http.(*connReader).backgroundRead(0xc0003d2090) net/http/server.go:690 +0x37 fp=0xc001725fc8 sp=0xc001725f80 pc=0x7ff68681e117 net/http.(*connReader).startBackgroundRead.gowrap2() net/http/server.go:686 +0x25 fp=0xc001725fe0 sp=0xc001725fc8 pc=0x7ff68681e045 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc001725fe8 sp=0xc001725fe0 pc=0x7ff68650db21 created by net/http.(*connReader).startBackgroundRead in goroutine 15 net/http/server.go:686 +0xb6

goroutine 1032 gp=0xc00098c380 m=nil [chan receive]: runtime.gopark(0x30?, 0x7ff687c8ea20?, 0x1?, 0x9d?, 0xc00041bb20?) runtime/proc.go:435 +0xce fp=0xc00041bad8 sp=0xc00041bab8 pc=0x7ff68650598e runtime.chanrecv(0xc000690230, 0x0, 0x1) runtime/chan.go:664 +0x445 fp=0xc00041bb50 sp=0xc00041bad8 pc=0x7ff6864a2d45 runtime.chanrecv1(0x7ff687d96d37?, 0x2c?) runtime/chan.go:506 +0x12 fp=0xc00041bb78 sp=0xc00041bb50 pc=0x7ff6864a28d2 github.com/ollama/ollama/runner/ollamarunner.(*Server).computeBatch(0xc00051c000, {0x1, {0x7ff687fb3400, 0xc009726000}, {0x7ff687fc3298, 0xc000589d70}, {0xc0009b6008, 0x200, 0x25f}, {{0x7ff687fc3298, ...}, ...}, ...}) github.com/ollama/ollama/runner/ollamarunner/runner.go:645 +0x185 fp=0xc00041bef0 sp=0xc00041bb78 pc=0x7ff686ba0f05 github.com/ollama/ollama/runner/ollamarunner.(*Server).run.gowrap1() github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x58 fp=0xc00041bfe0 sp=0xc00041bef0 pc=0x7ff686b9f2d8 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc00041bfe8 sp=0xc00041bfe0 pc=0x7ff68650db21 created by github.com/ollama/ollama/runner/ollamarunner.(*Server).run in goroutine 50 github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x2cd rax 0x0 rbx 0x3f79aff628 rcx 0x700000006 rdx 0x0 rdi 0xe06d7363 rsi 0x1 rbp 0x4 rsp 0x3f79aff500 r8 0x1 r9 0xe06d7363 r10 0x0 r11 0x1dfb20d4980 r12 0x0 r13 0x1e0458e5e90 r14 0x8000 r15 0x0 rip 0x7ffe3ffd79da rflags 0x206 cs 0x33 fs 0x53 gs 0x2b time=2026-04-30T11:45:18.251+08:00 level=ERROR source=server.go:1624 msg="post predict" error="Post "http://127.0.0.1:3238/completion\": read tcp 127.0.0.1:3245->127.0.0.1:3238: wsarecv: An existing connection was forcibly closed by the remote host." [GIN] 2026/04/30 - 11:45:18 | 500 | 16.2620155s | 127.0.0.1 | POST "/v1/chat/completions" time=2026-04-30T11:45:19.380+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\Users\Administrator\AppData\Local\Programs\Ollama\ollama.exe runner --ollama-engine --port 3297" time=2026-04-30T11:45:20.377+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\Users\Administrator\AppData\Local\Programs\Ollama\ollama.exe runner --ollama-engine --port 3316" time=2026-04-30T11:45:21.071+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\Users\Administrator\AppData\Local\Programs\Ollama\ollama.exe runner --ollama-engine --port 3331" time=2026-04-30T11:45:21.775+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\Users\Administrator\AppData\Local\Programs\Ollama\ollama.exe runner --ollama-engine --port 3346" time=2026-04-30T11:45:22.457+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\Users\Administrator\AppData\Local\Programs\Ollama\ollama.exe runner --ollama-engine --port 3357" time=2026-04-30T11:45:23.173+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\Users\Administrator\AppData\Local\Programs\Ollama\ollama.exe runner --ollama-engine --port 3377" time=2026-04-30T11:45:23.828+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\Users\Administrator\AppData\Local\Programs\Ollama\ollama.exe runner --ollama-engine --port 3388" time=2026-04-30T11:45:24.528+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\Users\Administrator\AppData\Local\Programs\Ollama\ollama.exe runner --ollama-engine --port 3400" time=2026-04-30T11:45:25.125+08:00 level=INFO source=runner.go:464 msg="failure during GPU discovery" OLLAMA_LIBRARY_PATH="[C:\Users\Administrator\AppData\Local\Programs\Ollama\lib\ollama C:\Users\Administrator\AppData\Local\Programs\Ollama\lib\ollama\vulkan]" extra_envs=map[] error="failed to finish discovery before timeout" time=2026-04-30T11:45:25.125+08:00 level=WARN source=runner.go:356 msg="unable to refresh free memory, using old values" time=2026-04-30T11:45:25.127+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\Users\Administrator\AppData\Local\Programs\Ollama\ollama.exe runner --ollama-engine --port 3408" time=2026-04-30T11:45:25.772+08:00 level=INFO source=cpu_windows.go:148 msg=packages count=1 time=2026-04-30T11:45:25.772+08:00 level=INFO source=cpu_windows.go:164 msg="efficiency cores detected" maxEfficiencyClass=1 time=2026-04-30T11:45:25.772+08:00 level=INFO source=cpu_windows.go:195 msg="" package=0 cores=10 efficiency=4 threads=16 time=2026-04-30T11:45:26.023+08:00 level=INFO source=model.go:97 msg="gemma4: token IDs" image=255999 image_end=258882 audio=256000 audio_end=258883 time=2026-04-30T11:45:26.024+08:00 level=INFO source=server.go:259 msg="enabling flash attention" time=2026-04-30T11:45:26.025+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\Users\Administrator\AppData\Local\Programs\Ollama\ollama.exe runner --ollama-engine --model D:\AI_MODELS\models\blobs\sha256-7121486771cbfe218851513210c40b35dbdee93ab1ef43fe36283c883980f0df --port 3427" time=2026-04-30T11:45:26.053+08:00 level=INFO source=sched.go:484 msg="system memory" total="47.7 GiB" free="31.5 GiB" free_swap="46.7 GiB" time=2026-04-30T11:45:26.053+08:00 level=INFO source=sched.go:491 msg="gpu memory" id=8680a156-0800-0000-0300-000000000000 library=Vulkan available="4.9 GiB" free="5.4 GiB" minimum="457.0 MiB" overhead="0 B" time=2026-04-30T11:45:26.053+08:00 level=INFO source=server.go:771 msg="loading model" "model layers"=31 requested=-1 time=2026-04-30T11:45:26.103+08:00 level=INFO source=runner.go:1417 msg="starting ollama engine" time=2026-04-30T11:45:26.111+08:00 level=INFO source=runner.go:1452 msg="Server listening on 127.0.0.1:3427" time=2026-04-30T11:45:26.121+08:00 level=INFO source=runner.go:1290 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:31[ID:8680a156-0800-0000-0300-000000000000 Layers:31(0..30)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" time=2026-04-30T11:45:26.174+08:00 level=INFO source=ggml.go:136 msg="" architecture=gemma4 file_type=Q4_K_M name="" description="" num_tensors=1014 num_key_values=52 load_backend: loaded CPU backend from C:\Users\Administrator\AppData\Local\Programs\Ollama\lib\ollama\ggml-cpu-alderlake.dll ggml_vulkan: Found 1 Vulkan devices: ggml_vulkan: 0 = Intel(R) Arc(TM) A750 Graphics (Intel Corporation) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: none load_backend: loaded Vulkan backend from C:\Users\Administrator\AppData\Local\Programs\Ollama\lib\ollama\vulkan\ggml-vulkan.dll time=2026-04-30T11:45:26.288+08:00 level=INFO source=ggml.go:104 msg=system CPU.0.SSE3=1 CPU.0.SSSE3=1 CPU.0.AVX=1 CPU.0.AVX_VNNI=1 CPU.0.AVX2=1 CPU.0.F16C=1 CPU.0.FMA=1 CPU.0.BMI2=1 CPU.0.LLAMAFILE=1 CPU.1.LLAMAFILE=1 compiler=cgo(clang) ggml_backend_vk_get_device_memory called: uuid 8680a156-0800-0000-0300-000000000000 ggml_backend_vk_get_device_memory called: luid 0x000000000000f6b6 ggml_dxgi_pdh_init called DXGI + PDH Initialized. Getting GPU free memory info [DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000000F6B6, Dedicated: 7.91 GB, Shared: 23.87 GB [DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000001C965, Dedicated: 7.91 GB, Shared: 23.87 GB [DXGI] Adapter Description: Microsoft Basic Render Driver, LUID: 0x000000000000FAF4, Dedicated: 0.00 GB, Shared: 23.87 GB Discrete GPU (Intel(R) Arc(TM) A750 Graphics) with LUID 0x000000000000f6b6 detected. Dedicated Total: 8491368448.00 bytes (7.91 GB), Dedicated Usage: 2732904448.00 bytes (2.55 GB) ggml_backend_vk_get_device_memory utilizing DXGI + PDH memory reporting free: 5758464000 total: 8491368448 time=2026-04-30T11:45:26.533+08:00 level=INFO source=model.go:97 msg="gemma4: token IDs" image=255999 image_end=258882 audio=256000 audio_end=258883 time=2026-04-30T11:45:26.547+08:00 level=INFO source=model.go:138 msg="vision: decode" elapsed=2.1271ms bounds=(0,0)-(2048,2048) time=2026-04-30T11:45:26.648+08:00 level=INFO source=model.go:145 msg="vision: preprocess" elapsed=101.1927ms size="[768 768]" time=2026-04-30T11:45:26.648+08:00 level=INFO source=model.go:148 msg="vision: pixelValues" shape="[768 768 3]" dim0=768 dim1=768 dim2=3 time=2026-04-30T11:45:26.648+08:00 level=INFO source=model.go:152 msg="vision: patches" patchesX=48 patchesY=48 total=2304 patchSize=16 time=2026-04-30T11:45:26.649+08:00 level=INFO source=model.go:156 msg="vision: encoded" elapsed=104.8169ms shape="[2816 256]" time=2026-04-30T11:45:26.675+08:00 level=INFO source=runner.go:1290 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:8[ID:8680a156-0800-0000-0300-000000000000 Layers:8(22..29)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" ggml_backend_vk_get_device_memory called: uuid 8680a156-0800-0000-0300-000000000000 ggml_backend_vk_get_device_memory called: luid 0x000000000000f6b6 ggml_dxgi_pdh_init called DXGI + PDH Initialized. Getting GPU free memory info [DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000000F6B6, Dedicated: 7.91 GB, Shared: 23.87 GB [DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000001C965, Dedicated: 7.91 GB, Shared: 23.87 GB [DXGI] Adapter Description: Microsoft Basic Render Driver, LUID: 0x000000000000FAF4, Dedicated: 0.00 GB, Shared: 23.87 GB Discrete GPU (Intel(R) Arc(TM) A750 Graphics) with LUID 0x000000000000f6b6 detected. Dedicated Total: 8491368448.00 bytes (7.91 GB), Dedicated Usage: 2732904448.00 bytes (2.55 GB) ggml_backend_vk_get_device_memory utilizing DXGI + PDH memory reporting free: 5758464000 total: 8491368448 time=2026-04-30T11:45:26.809+08:00 level=INFO source=model.go:97 msg="gemma4: token IDs" image=255999 image_end=258882 audio=256000 audio_end=258883 time=2026-04-30T11:45:26.823+08:00 level=INFO source=model.go:138 msg="vision: decode" elapsed=1.0667ms bounds=(0,0)-(2048,2048) time=2026-04-30T11:45:26.920+08:00 level=INFO source=model.go:145 msg="vision: preprocess" elapsed=97.1047ms size="[768 768]" time=2026-04-30T11:45:26.920+08:00 level=INFO source=model.go:148 msg="vision: pixelValues" shape="[768 768 3]" dim0=768 dim1=768 dim2=3 time=2026-04-30T11:45:26.920+08:00 level=INFO source=model.go:152 msg="vision: patches" patchesX=48 patchesY=48 total=2304 patchSize=16 time=2026-04-30T11:45:26.921+08:00 level=INFO source=model.go:156 msg="vision: encoded" elapsed=99.0688ms shape="[2816 256]" time=2026-04-30T11:45:26.947+08:00 level=INFO source=runner.go:1290 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:8[ID:8680a156-0800-0000-0300-000000000000 Layers:8(22..29)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" ggml_backend_vk_get_device_memory called: uuid 8680a156-0800-0000-0300-000000000000 ggml_backend_vk_get_device_memory called: luid 0x000000000000f6b6 ggml_dxgi_pdh_init called DXGI + PDH Initialized. Getting GPU free memory info [DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000000F6B6, Dedicated: 7.91 GB, Shared: 23.87 GB [DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000001C965, Dedicated: 7.91 GB, Shared: 23.87 GB [DXGI] Adapter Description: Microsoft Basic Render Driver, LUID: 0x000000000000FAF4, Dedicated: 0.00 GB, Shared: 23.87 GB Discrete GPU (Intel(R) Arc(TM) A750 Graphics) with LUID 0x000000000000f6b6 detected. Dedicated Total: 8491368448.00 bytes (7.91 GB), Dedicated Usage: 2732904448.00 bytes (2.55 GB) ggml_backend_vk_get_device_memory utilizing DXGI + PDH memory reporting free: 5758464000 total: 8491368448 time=2026-04-30T11:45:27.352+08:00 level=INFO source=model.go:97 msg="gemma4: token IDs" image=255999 image_end=258882 audio=256000 audio_end=258883 time=2026-04-30T11:45:27.361+08:00 level=INFO source=model.go:138 msg="vision: decode" elapsed=1.049ms bounds=(0,0)-(2048,2048) time=2026-04-30T11:45:27.461+08:00 level=INFO source=model.go:145 msg="vision: preprocess" elapsed=99.9253ms size="[768 768]" time=2026-04-30T11:45:27.463+08:00 level=INFO source=model.go:148 msg="vision: pixelValues" shape="[768 768 3]" dim0=768 dim1=768 dim2=3 time=2026-04-30T11:45:27.463+08:00 level=INFO source=model.go:152 msg="vision: patches" patchesX=48 patchesY=48 total=2304 patchSize=16 time=2026-04-30T11:45:27.464+08:00 level=INFO source=model.go:156 msg="vision: encoded" elapsed=103.7039ms shape="[2816 256]" time=2026-04-30T11:45:27.690+08:00 level=INFO source=runner.go:1290 msg=load request="{Operation:commit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:8[ID:8680a156-0800-0000-0300-000000000000 Layers:8(22..29)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" time=2026-04-30T11:45:27.690+08:00 level=INFO source=ggml.go:482 msg="offloading 8 repeating layers to GPU" time=2026-04-30T11:45:27.690+08:00 level=INFO source=ggml.go:486 msg="offloading output layer to CPU" time=2026-04-30T11:45:27.690+08:00 level=INFO source=ggml.go:494 msg="offloaded 8/31 layers to GPU" time=2026-04-30T11:45:27.690+08:00 level=INFO source=device.go:240 msg="model weights" device=Vulkan0 size="4.1 GiB" time=2026-04-30T11:45:27.690+08:00 level=INFO source=device.go:245 msg="model weights" device=CPU size="13.2 GiB" time=2026-04-30T11:45:27.690+08:00 level=INFO source=device.go:251 msg="kv cache" device=Vulkan0 size="224.0 MiB" time=2026-04-30T11:45:27.690+08:00 level=INFO source=device.go:256 msg="kv cache" device=CPU size="656.0 MiB" time=2026-04-30T11:45:27.691+08:00 level=INFO source=device.go:262 msg="compute graph" device=Vulkan0 size="352.3 MiB" time=2026-04-30T11:45:27.691+08:00 level=INFO source=device.go:267 msg="compute graph" device=CPU size="55.0 MiB" time=2026-04-30T11:45:27.692+08:00 level=INFO source=device.go:272 msg="total memory" size="18.6 GiB" time=2026-04-30T11:45:27.692+08:00 level=INFO source=sched.go:561 msg="loaded runners" count=1 time=2026-04-30T11:45:27.693+08:00 level=INFO source=server.go:1364 msg="waiting for llama runner to start responding" time=2026-04-30T11:45:27.695+08:00 level=INFO source=server.go:1398 msg="waiting for server to become available" status="llm server loading model" time=2026-04-30T11:45:31.144+08:00 level=INFO source=server.go:1402 msg="llama runner started in 5.09 seconds" Exception 0xe06d7363 0x19930520 0x4b92aff500 0x7ffe3ffd79da PC=0x7ffe3ffd79da signal arrived during external code execution

runtime.cgocall(0x7ff687680aa0, 0xc000f0daa0) runtime/cgocall.go:167 +0x3e fp=0xc000f0da78 sp=0xc000f0da10 pc=0x7ff68650243e github.com/ollama/ollama/ml/backend/ggml._Cfunc_ggml_backend_sched_graph_compute_async(0x17ac69a07c0, 0x17f45d3f600) _cgo_gotypes.go:963 +0x50 fp=0xc000f0daa0 sp=0xc000f0da78 pc=0x7ff686a42370 github.com/ollama/ollama/ml/backend/ggml.(*Context).ComputeWithNotify.func2(...) github.com/ollama/ollama/ml/backend/ggml/ggml.go:825 github.com/ollama/ollama/ml/backend/ggml.(*Context).ComputeWithNotify(0xc00972c840, 0xc0002db800?, {0xc000591950, 0x1, 0x2?}) github.com/ollama/ollama/ml/backend/ggml/ggml.go:825 +0x1b5 fp=0xc000f0db78 sp=0xc000f0daa0 pc=0x7ff686a520b5 github.com/ollama/ollama/runner/ollamarunner.(*Server).computeBatch(0xc0004c6000, {0x0, {0x7ff687fb3400, 0xc00972c840}, {0x7ff687fc3298, 0xc0097e2de0}, {0xc001d0c608, 0x200, 0x25f}, {{0x7ff687fc3298, ...}, ...}, ...}) github.com/ollama/ollama/runner/ollamarunner/runner.go:716 +0x862 fp=0xc000f0def0 sp=0xc000f0db78 pc=0x7ff686ba15e2 github.com/ollama/ollama/runner/ollamarunner.(*Server).run.gowrap1() github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x58 fp=0xc000f0dfe0 sp=0xc000f0def0 pc=0x7ff686b9f2d8 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc000f0dfe8 sp=0xc000f0dfe0 pc=0x7ff68650db21 created by github.com/ollama/ollama/runner/ollamarunner.(*Server).run in goroutine 37 github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x2cd

goroutine 1 gp=0xc0000021c0 m=nil [IO wait, locked to thread]: runtime.gopark(0x7ff68650f320?, 0x7ff688b5c280?, 0x20?, 0x80?, 0xc0004c80cc?) runtime/proc.go:435 +0xce fp=0xc001685630 sp=0xc001685610 pc=0x7ff68650598e runtime.netpollblock(0x26c?, 0x864a0406?, 0xf6?) runtime/netpoll.go:575 +0xf7 fp=0xc001685668 sp=0xc001685630 pc=0x7ff6864cbdf7 internal/poll.runtime_pollWait(0x17a78d7e8f0, 0x72) runtime/netpoll.go:351 +0x85 fp=0xc001685688 sp=0xc001685668 pc=0x7ff686504b25 internal/poll.(*pollDesc).wait(0x7ff68659aad3?, 0x0?, 0x0) internal/poll/fd_poll_runtime.go:84 +0x27 fp=0xc0016856b0 sp=0xc001685688 pc=0x7ff68659c0c7 internal/poll.execIO(0xc0004c8020, 0xc0003a7758) internal/poll/fd_windows.go:177 +0x105 fp=0xc001685728 sp=0xc0016856b0 pc=0x7ff68659d525 internal/poll.(*FD).acceptOne(0xc0004c8008, 0x2b8, {0xc0004d40f0?, 0xc0003a77b8?, 0x7ff6865a51e5?}, 0xc0003a77ec?) internal/poll/fd_windows.go:946 +0x65 fp=0xc001685788 sp=0xc001685728 pc=0x7ff6865a1aa5 internal/poll.(*FD).Accept(0xc0004c8008, 0xc001685938) internal/poll/fd_windows.go:980 +0x1b6 fp=0xc001685840 sp=0xc001685788 pc=0x7ff6865a1dd6 net.(*netFD).accept(0xc0004c8008) net/fd_windows.go:182 +0x4b fp=0xc001685958 sp=0xc001685840 pc=0x7ff686613a6b net.(*TCPListener).accept(0xc0001c4080) net/tcpsock_posix.go:159 +0x1b fp=0xc0016859a8 sp=0xc001685958 pc=0x7ff68662a01b net.(*TCPListener).Accept(0xc0001c4080) net/tcpsock.go:380 +0x30 fp=0xc0016859d8 sp=0xc0016859a8 pc=0x7ff686628dd0 net/http.(*onceCloseListener).Accept(0xc0004d2090?) <autogenerated>:1 +0x24 fp=0xc0016859f0 sp=0xc0016859d8 pc=0x7ff686851e44 net/http.(*Server).Serve(0xc000690400, {0x7ff687fa2e70, 0xc0001c4080}) net/http/server.go:3424 +0x30c fp=0xc001685b20 sp=0xc0016859f0 pc=0x7ff68682970c github.com/ollama/ollama/runner/ollamarunner.Execute({0xc00004c0b0, 0x4, 0x5}) github.com/ollama/ollama/runner/ollamarunner/runner.go:1453 +0x94e fp=0xc001685cf0 sp=0xc001685b20 pc=0x7ff686ba866e github.com/ollama/ollama/runner.Execute({0xc00004c090?, 0x0?, 0x0?}) github.com/ollama/ollama/runner/runner.go:18 +0x10e fp=0xc001685d30 sp=0xc001685cf0 pc=0x7ff686c6d0ae github.com/ollama/ollama/cmd.NewCLI.func3(0xc00014d800?, {0x7ff687d47a56?, 0x4?, 0x7ff687d47a5a?}) github.com/ollama/ollama/cmd/cmd.go:2327 +0x45 fp=0xc001685d58 sp=0xc001685d30 pc=0x7ff6874dbaa5 github.com/spf13/cobra.(*Command).execute(0xc00047bb08, {0xc000696d70, 0x5, 0x5}) github.com/spf13/[email protected]/command.go:940 +0x85c fp=0xc001685e78 sp=0xc001685d58 pc=0x7ff68668ec3c github.com/spf13/cobra.(*Command).ExecuteC(0xc000420908) github.com/spf13/[email protected]/command.go:1068 +0x3a5 fp=0xc001685f30 sp=0xc001685e78 pc=0x7ff68668f485 github.com/spf13/cobra.(*Command).Execute(...) github.com/spf13/[email protected]/command.go:992 github.com/spf13/cobra.(*Command).ExecuteContext(...) github.com/spf13/[email protected]/command.go:985 main.main() github.com/ollama/ollama/main.go:12 +0x4d fp=0xc001685f50 sp=0xc001685f30 pc=0x7ff6874dd42d runtime.main() runtime/proc.go:283 +0x27d fp=0xc001685fe0 sp=0xc001685f50 pc=0x7ff6864d4ddd runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc001685fe8 sp=0xc001685fe0 pc=0x7ff68650db21

goroutine 2 gp=0xc0000028c0 m=nil [force gc (idle)]: runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?) runtime/proc.go:435 +0xce fp=0xc000079fa8 sp=0xc000079f88 pc=0x7ff68650598e runtime.goparkunlock(...) runtime/proc.go:441 runtime.forcegchelper() runtime/proc.go:348 +0xb8 fp=0xc000079fe0 sp=0xc000079fa8 pc=0x7ff6864d50f8 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc000079fe8 sp=0xc000079fe0 pc=0x7ff68650db21 created by runtime.init.7 in goroutine 1 runtime/proc.go:336 +0x1a

goroutine 3 gp=0xc000002c40 m=nil [GC sweep wait]: runtime.gopark(0x1?, 0x0?, 0x0?, 0x0?, 0x0?) runtime/proc.go:435 +0xce fp=0xc00007bf80 sp=0xc00007bf60 pc=0x7ff68650598e runtime.goparkunlock(...) runtime/proc.go:441 runtime.bgsweep(0xc000088000) runtime/mgcsweep.go:316 +0xdf fp=0xc00007bfc8 sp=0xc00007bf80 pc=0x7ff6864bdebf runtime.gcenable.gowrap1() runtime/mgc.go:204 +0x25 fp=0xc00007bfe0 sp=0xc00007bfc8 pc=0x7ff6864b2285 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc00007bfe8 sp=0xc00007bfe0 pc=0x7ff68650db21 created by runtime.gcenable in goroutine 1 runtime/mgc.go:204 +0x66

goroutine 4 gp=0xc000002e00 m=nil [GC scavenge wait]: runtime.gopark(0x277c34?, 0x7267d5?, 0x0?, 0x0?, 0x0?) runtime/proc.go:435 +0xce fp=0xc00008ff78 sp=0xc00008ff58 pc=0x7ff68650598e runtime.goparkunlock(...) runtime/proc.go:441 runtime.(*scavengerState).park(0x7ff688b8a740) runtime/mgcscavenge.go:425 +0x49 fp=0xc00008ffa8 sp=0xc00008ff78 pc=0x7ff6864bb909 runtime.bgscavenge(0xc000088000) runtime/mgcscavenge.go:658 +0x59 fp=0xc00008ffc8 sp=0xc00008ffa8 pc=0x7ff6864bbe99 runtime.gcenable.gowrap2() runtime/mgc.go:205 +0x25 fp=0xc00008ffe0 sp=0xc00008ffc8 pc=0x7ff6864b2225 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc00008ffe8 sp=0xc00008ffe0 pc=0x7ff68650db21 created by runtime.gcenable in goroutine 1 runtime/mgc.go:205 +0xa5

goroutine 5 gp=0xc000003340 m=nil [finalizer wait]: runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?) runtime/proc.go:435 +0xce fp=0xc000091e30 sp=0xc000091e10 pc=0x7ff68650598e runtime.runfinq() runtime/mfinal.go:196 +0x107 fp=0xc000091fe0 sp=0xc000091e30 pc=0x7ff6864b1207 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc000091fe8 sp=0xc000091fe0 pc=0x7ff68650db21 created by runtime.createfing in goroutine 1 runtime/mfinal.go:166 +0x3d

goroutine 6 gp=0xc000003dc0 m=nil [chan receive]: runtime.gopark(0xc0000e1900?, 0xc0097e2e28?, 0x60?, 0xdf?, 0x7ff6865fc688?) runtime/proc.go:435 +0xce fp=0xc00007df18 sp=0xc00007def8 pc=0x7ff68650598e runtime.chanrecv(0xc00003a5b0, 0x0, 0x1) runtime/chan.go:664 +0x445 fp=0xc00007df90 sp=0xc00007df18 pc=0x7ff6864a2d45 runtime.chanrecv1(0x7ff6864d4f40?, 0xc00007df76?) runtime/chan.go:506 +0x12 fp=0xc00007dfb8 sp=0xc00007df90 pc=0x7ff6864a28d2 runtime.unique_runtime_registerUniqueMapCleanup.func2(...) runtime/mgc.go:1796 runtime.unique_runtime_registerUniqueMapCleanup.gowrap1() runtime/mgc.go:1799 +0x2f fp=0xc00007dfe0 sp=0xc00007dfb8 pc=0x7ff6864b54af runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc00007dfe8 sp=0xc00007dfe0 pc=0x7ff68650db21 created by unique.runtime_registerUniqueMapCleanup in goroutine 1 runtime/mgc.go:1794 +0x85

goroutine 18 gp=0xc0001f8000 m=nil [GC worker (idle)]: runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?) runtime/proc.go:435 +0xce fp=0xc00008bf38 sp=0xc00008bf18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc00008bfc8 sp=0xc00008bf38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc00008bfe0 sp=0xc00008bfc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc00008bfe8 sp=0xc00008bfe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 7 gp=0xc0004021c0 m=nil [GC worker (idle)]: runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?) runtime/proc.go:435 +0xce fp=0xc000409f38 sp=0xc000409f18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc000409fc8 sp=0xc000409f38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc000409fe0 sp=0xc000409fc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc000409fe8 sp=0xc000409fe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 8 gp=0xc000402380 m=nil [GC worker (idle)]: runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?) runtime/proc.go:435 +0xce fp=0xc00040bf38 sp=0xc00040bf18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc00040bfc8 sp=0xc00040bf38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc00040bfe0 sp=0xc00040bfc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc00040bfe8 sp=0xc00040bfe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 9 gp=0xc000402540 m=nil [GC worker (idle)]: runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?) runtime/proc.go:435 +0xce fp=0xc000405f38 sp=0xc000405f18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc000405fc8 sp=0xc000405f38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc000405fe0 sp=0xc000405fc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc000405fe8 sp=0xc000405fe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 10 gp=0xc000402700 m=nil [GC worker (idle)]: runtime.gopark(0x57dd78c1be4?, 0x0?, 0x0?, 0x0?, 0x0?) runtime/proc.go:435 +0xce fp=0xc000407f38 sp=0xc000407f18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc000407fc8 sp=0xc000407f38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc000407fe0 sp=0xc000407fc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc000407fe8 sp=0xc000407fe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 11 gp=0xc0004028c0 m=nil [GC worker (idle)]: runtime.gopark(0x57ce6b85278?, 0x0?, 0x0?, 0x0?, 0x0?) runtime/proc.go:435 +0xce fp=0xc000411f38 sp=0xc000411f18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc000411fc8 sp=0xc000411f38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc000411fe0 sp=0xc000411fc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc000411fe8 sp=0xc000411fe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 34 gp=0xc000484000 m=nil [GC worker (idle)]: runtime.gopark(0x7ff688bdd9c0?, 0x1?, 0x40?, 0x1b?, 0x0?) runtime/proc.go:435 +0xce fp=0xc00040df38 sp=0xc00040df18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc00040dfc8 sp=0xc00040df38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc00040dfe0 sp=0xc00040dfc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc00040dfe8 sp=0xc00040dfe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 19 gp=0xc0001f81c0 m=nil [GC worker (idle)]: runtime.gopark(0x57dd78c1be4?, 0x3?, 0x0?, 0x0?, 0x0?) runtime/proc.go:435 +0xce fp=0xc00008df38 sp=0xc00008df18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc00008dfc8 sp=0xc00008df38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc00008dfe0 sp=0xc00008dfc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc00008dfe8 sp=0xc00008dfe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 12 gp=0xc000402a80 m=nil [GC worker (idle)]: runtime.gopark(0x57dd78c1be4?, 0x1?, 0xcc?, 0x2?, 0x0?) runtime/proc.go:435 +0xce fp=0xc000413f38 sp=0xc000413f18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc000413fc8 sp=0xc000413f38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc000413fe0 sp=0xc000413fc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc000413fe8 sp=0xc000413fe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 13 gp=0xc000402c40 m=nil [GC worker (idle)]: runtime.gopark(0x57ce6b85278?, 0x1?, 0x88?, 0x2?, 0x0?) runtime/proc.go:435 +0xce fp=0xc000419f38 sp=0xc000419f18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc000419fc8 sp=0xc000419f38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc000419fe0 sp=0xc000419fc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc000419fe8 sp=0xc000419fe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 35 gp=0xc0004841c0 m=nil [GC worker (idle)]: runtime.gopark(0x57ce6b85278?, 0x1?, 0xa8?, 0x1c?, 0x0?) runtime/proc.go:435 +0xce fp=0xc00040ff38 sp=0xc00040ff18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc00040ffc8 sp=0xc00040ff38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc00040ffe0 sp=0xc00040ffc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc00040ffe8 sp=0xc00040ffe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 20 gp=0xc0001f8380 m=nil [GC worker (idle)]: runtime.gopark(0x57ce6b85278?, 0x3?, 0x0?, 0x0?, 0x0?) runtime/proc.go:435 +0xce fp=0xc000415f38 sp=0xc000415f18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc000415fc8 sp=0xc000415f38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc000415fe0 sp=0xc000415fc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc000415fe8 sp=0xc000415fe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 14 gp=0xc000402e00 m=nil [GC worker (idle)]: runtime.gopark(0x57dd78c1be4?, 0x1?, 0x3c?, 0x31?, 0x0?) runtime/proc.go:435 +0xce fp=0xc00041bf38 sp=0xc00041bf18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc00041bfc8 sp=0xc00041bf38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc00041bfe0 sp=0xc00041bfc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc00041bfe8 sp=0xc00041bfe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 36 gp=0xc000484380 m=nil [GC worker (idle)]: runtime.gopark(0x57dd8158f00?, 0x1?, 0x0?, 0x1a?, 0x0?) runtime/proc.go:435 +0xce fp=0xc00048bf38 sp=0xc00048bf18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc00048bfc8 sp=0xc00048bf38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc00048bfe0 sp=0xc00048bfc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc00048bfe8 sp=0xc00048bfe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 21 gp=0xc0001f8540 m=nil [GC worker (idle)]: runtime.gopark(0x57dd77cbf78?, 0x3?, 0x0?, 0x0?, 0x0?) runtime/proc.go:435 +0xce fp=0xc000417f38 sp=0xc000417f18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc000417fc8 sp=0xc000417f38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc000417fe0 sp=0xc000417fc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc000417fe8 sp=0xc000417fe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 15 gp=0xc000402fc0 m=nil [GC worker (idle)]: runtime.gopark(0x7ff688bdd9c0?, 0x1?, 0x28?, 0x2e?, 0x0?) runtime/proc.go:435 +0xce fp=0xc000487f38 sp=0xc000487f18 pc=0x7ff68650598e runtime.gcBgMarkWorker(0xc0003211f0) runtime/mgc.go:1423 +0xe9 fp=0xc000487fc8 sp=0xc000487f38 pc=0x7ff6864b47a9 runtime.gcBgMarkStartWorkers.gowrap1() runtime/mgc.go:1339 +0x25 fp=0xc000487fe0 sp=0xc000487fc8 pc=0x7ff6864b4685 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc000487fe8 sp=0xc000487fe0 pc=0x7ff68650db21 created by runtime.gcBgMarkStartWorkers in goroutine 1 runtime/mgc.go:1339 +0x105

goroutine 37 gp=0xc0001f88c0 m=nil [chan receive]: runtime.gopark(0x30?, 0xffffffffffffffff?, 0x3?, 0x0?, 0xc001fa7798?) runtime/proc.go:435 +0xce fp=0xc001fa7750 sp=0xc001fa7730 pc=0x7ff68650598e runtime.chanrecv(0xc00003a230, 0x0, 0x1) runtime/chan.go:664 +0x445 fp=0xc001fa77c8 sp=0xc001fa7750 pc=0x7ff6864a2d45 runtime.chanrecv1(0x7ff687d92ab6?, 0x29?) runtime/chan.go:506 +0x12 fp=0xc001fa77f0 sp=0xc001fa77c8 pc=0x7ff6864a28d2 github.com/ollama/ollama/runner/ollamarunner.(*Server).forwardBatch(_, {0x1, {0x7ff687fb3400, 0xc0096f8340}, {0x7ff687fc3298, 0xc0097c64f8}, {0xc00096b308, 0x200, 0x25f}, {{0x7ff687fc3298, ...}, ...}, ...}) github.com/ollama/ollama/runner/ollamarunner/runner.go:476 +0xfa fp=0xc001fa7b58 sp=0xc001fa77f0 pc=0x7ff686b9f3fa github.com/ollama/ollama/runner/ollamarunner.(*Server).run(0xc0004c6000, {0x7ff687fa59b0, 0xc0004c4000}) github.com/ollama/ollama/runner/ollamarunner/runner.go:453 +0x18c fp=0xc001fa7fb8 sp=0xc001fa7b58 pc=0x7ff686b9f0ac github.com/ollama/ollama/runner/ollamarunner.Execute.gowrap1() github.com/ollama/ollama/runner/ollamarunner/runner.go:1430 +0x28 fp=0xc001fa7fe0 sp=0xc001fa7fb8 pc=0x7ff686ba88e8 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc001fa7fe8 sp=0xc001fa7fe0 pc=0x7ff68650db21 created by github.com/ollama/ollama/runner/ollamarunner.Execute in goroutine 1 github.com/ollama/ollama/runner/ollamarunner/runner.go:1430 +0x4c9

goroutine 38 gp=0xc0001f8a80 m=nil [select]: runtime.gopark(0xc000047a08?, 0x2?, 0xc0?, 0x77?, 0xc00004786c?) runtime/proc.go:435 +0xce fp=0xc001687698 sp=0xc001687678 pc=0x7ff68650598e runtime.selectgo(0xc001687a08, 0xc000047868, 0xe3c?, 0x0, 0x1?, 0x1) runtime/select.go:351 +0x837 fp=0xc0016877d0 sp=0xc001687698 pc=0x7ff6864e6437 github.com/ollama/ollama/runner/ollamarunner.(*Server).completion(0xc0004c6000, {0x7ff687fa3020, 0xc0002d01c0}, 0xc0003de3c0) github.com/ollama/ollama/runner/ollamarunner/runner.go:956 +0xc4e fp=0xc001687ac0 sp=0xc0016877d0 pc=0x7ff686ba388e github.com/ollama/ollama/runner/ollamarunner.(*Server).completion-fm({0x7ff687fa3020?, 0xc0002d01c0?}, 0xc000047b40?) <autogenerated>:1 +0x36 fp=0xc001687af0 sp=0xc001687ac0 pc=0x7ff686ba8dd6 net/http.HandlerFunc.ServeHTTP(0xc0004ca000?, {0x7ff687fa3020?, 0xc0002d01c0?}, 0xc000047b60?) net/http/server.go:2294 +0x29 fp=0xc001687b18 sp=0xc001687af0 pc=0x7ff686825d49 net/http.(*ServeMux).ServeHTTP(0x7ff6864ab785?, {0x7ff687fa3020, 0xc0002d01c0}, 0xc0003de3c0) net/http/server.go:2822 +0x1c4 fp=0xc001687b68 sp=0xc001687b18 pc=0x7ff686827c44 net/http.serverHandler.ServeHTTP({0x7ff687f9eeb0?}, {0x7ff687fa3020?, 0xc0002d01c0?}, 0x1?) net/http/server.go:3301 +0x8e fp=0xc001687b98 sp=0xc001687b68 pc=0x7ff6868456ce net/http.(*conn).serve(0xc0004d2090, {0x7ff687fa5978, 0xc0004c0360}) net/http/server.go:2102 +0x625 fp=0xc001687fb8 sp=0xc001687b98 pc=0x7ff686824245 net/http.(*Server).Serve.gowrap3() net/http/server.go:3454 +0x28 fp=0xc001687fe0 sp=0xc001687fb8 pc=0x7ff686829b08 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc001687fe8 sp=0xc001687fe0 pc=0x7ff68650db21 created by net/http.(*Server).Serve in goroutine 1 net/http/server.go:3454 +0x485

goroutine 1142 gp=0xc000485880 m=nil [IO wait]: runtime.gopark(0x0?, 0xc0004c82a0?, 0x48?, 0x83?, 0xc0004c834c?) runtime/proc.go:435 +0xce fp=0xc001097d58 sp=0xc001097d38 pc=0x7ff68650598e runtime.netpollblock(0x214?, 0x864a0406?, 0xf6?) runtime/netpoll.go:575 +0xf7 fp=0xc001097d90 sp=0xc001097d58 pc=0x7ff6864cbdf7 internal/poll.runtime_pollWait(0x17a78d7e7d8, 0x72) runtime/netpoll.go:351 +0x85 fp=0xc001097db0 sp=0xc001097d90 pc=0x7ff686504b25 internal/poll.(*pollDesc).wait(0x214?, 0x72?, 0x0) internal/poll/fd_poll_runtime.go:84 +0x27 fp=0xc001097dd8 sp=0xc001097db0 pc=0x7ff68659c0c7 internal/poll.execIO(0xc0004c82a0, 0x7ff687dd8778) internal/poll/fd_windows.go:177 +0x105 fp=0xc001097e50 sp=0xc001097dd8 pc=0x7ff68659d525 internal/poll.(*FD).Read(0xc0004c8288, {0xc000188041, 0x1, 0x1}) internal/poll/fd_windows.go:438 +0x29b fp=0xc001097ef0 sp=0xc001097e50 pc=0x7ff68659e1fb net.(*netFD).Read(0xc0004c8288, {0xc000188041?, 0xc00068e758?, 0xc001097f70?}) net/fd_posix.go:55 +0x25 fp=0xc001097f38 sp=0xc001097ef0 pc=0x7ff686611945 net.(*conn).Read(0xc00020c090, {0xc000188041?, 0x7ff687facb50?, 0xc00029e320?}) net/net.go:194 +0x45 fp=0xc001097f80 sp=0xc001097f38 pc=0x7ff686621065 net/http.(*connReader).backgroundRead(0xc000188030) net/http/server.go:690 +0x37 fp=0xc001097fc8 sp=0xc001097f80 pc=0x7ff68681e117 net/http.(*connReader).startBackgroundRead.gowrap2() net/http/server.go:686 +0x25 fp=0xc001097fe0 sp=0xc001097fc8 pc=0x7ff68681e045 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc001097fe8 sp=0xc001097fe0 pc=0x7ff68650db21 created by net/http.(*connReader).startBackgroundRead in goroutine 38 net/http/server.go:686 +0xb6

goroutine 1172 gp=0xc00058ea80 m=nil [chan receive]: runtime.gopark(0x30?, 0x247ff01?, 0x60?, 0x4?, 0xc0001adb20?) runtime/proc.go:435 +0xce fp=0xc0001adad8 sp=0xc0001adab8 pc=0x7ff68650598e runtime.chanrecv(0xc0003201c0, 0x0, 0x1) runtime/chan.go:664 +0x445 fp=0xc0001adb50 sp=0xc0001adad8 pc=0x7ff6864a2d45 runtime.chanrecv1(0x7ff687d96d37?, 0x2c?) runtime/chan.go:506 +0x12 fp=0xc0001adb78 sp=0xc0001adb50 pc=0x7ff6864a28d2 github.com/ollama/ollama/runner/ollamarunner.(*Server).computeBatch(0xc0004c6000, {0x1, {0x7ff687fb3400, 0xc0096f8340}, {0x7ff687fc3298, 0xc0097c64f8}, {0xc00096b308, 0x200, 0x25f}, {{0x7ff687fc3298, ...}, ...}, ...}) github.com/ollama/ollama/runner/ollamarunner/runner.go:645 +0x185 fp=0xc0001adef0 sp=0xc0001adb78 pc=0x7ff686ba0f05 github.com/ollama/ollama/runner/ollamarunner.(*Server).run.gowrap1() github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x58 fp=0xc0001adfe0 sp=0xc0001adef0 pc=0x7ff686b9f2d8 runtime.goexit({}) runtime/asm_amd64.s:1700 +0x1 fp=0xc0001adfe8 sp=0xc0001adfe0 pc=0x7ff68650db21 created by github.com/ollama/ollama/runner/ollamarunner.(*Server).run in goroutine 37 github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x2cd rax 0x0 rbx 0x4b92aff488 rcx 0x700000006 rdx 0x0 rdi 0xe06d7363 rsi 0x1 rbp 0x4 rsp 0x4b92aff360 r8 0x1 r9 0xe06d7363 r10 0x0 r11 0x17a732c2100 r12 0x0 r13 0x17ac69a07c0 r14 0x1000000 r15 0x0 rip 0x7ffe3ffd79da rflags 0x206 cs 0x33 fs 0x53 gs 0x2b time=2026-04-30T11:45:34.878+08:00 level=ERROR source=server.go:1624 msg="post predict" error="Post "http://127.0.0.1:3427/completion\": read tcp 127.0.0.1:3434->127.0.0.1:3427: wsarecv: An existing connection was forcibly closed by the remote host." [GIN] 2026/04/30 - 11:45:34 | 500 | 15.7829065s | 127.0.0.1 | POST "/v1/chat/completions"

Root Cause

The issue may be related to the earlier label leakage report (#15888) – possibly the same root cause manifesting as either label leakage or full crash depending on timing/memory state

Fix Action

Fix / Workaround

Workaround

Code Example

0ac
github.com/ollama/ollama/runner/ollamarunner.Execute.gowrap1()
        github.com/ollama/ollama/runner/ollamarunner/runner.go:1430 +0x28 fp=0xc00086ffe0 sp=0xc00086ffb8 pc=0x7ff686ba88e8
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00086ffe8 sp=0xc00086ffe0 pc=0x7ff68650db21
created by github.com/ollama/ollama/runner/ollamarunner.Execute in goroutine 1
        github.com/ollama/ollama/runner/ollamarunner/runner.go:1430 +0x4c9

goroutine 15 gp=0xc0004168c0 m=nil [select]:
runtime.gopark(0xc000047a08?, 0x2?, 0xc0?, 0x77?, 0xc00004786c?)
        runtime/proc.go:435 +0xce fp=0xc000047698 sp=0xc000047678 pc=0x7ff68650598e
runtime.selectgo(0xc000047a08, 0xc000047868, 0xe3c?, 0x0, 0x1?, 0x1)
        runtime/select.go:351 +0x837 fp=0xc0000477d0 sp=0xc000047698 pc=0x7ff6864e6437
github.com/ollama/ollama/runner/ollamarunner.(*Server).completion(0xc00051c000, {0x7ff687fa3020, 0xc000692000}, 0xc000616000)
        github.com/ollama/ollama/runner/ollamarunner/runner.go:956 +0xc4e fp=0xc000047ac0 sp=0xc0000477d0 pc=0x7ff686ba388e
github.com/ollama/ollama/runner/ollamarunner.(*Server).completion-fm({0x7ff687fa3020?, 0xc000692000?}, 0xc000047b40?)
        <autogenerated>:1 +0x36 fp=0xc000047af0 sp=0xc000047ac0 pc=0x7ff686ba8dd6
net/http.HandlerFunc.ServeHTTP(0xc000520000?, {0x7ff687fa3020?, 0xc000692000?}, 0xc000047b60?)
        net/http/server.go:2294 +0x29 fp=0xc000047b18 sp=0xc000047af0 pc=0x7ff686825d49
net/http.(*ServeMux).ServeHTTP(0x7ff6864ab785?, {0x7ff687fa3020, 0xc000692000}, 0xc000616000)
        net/http/server.go:2822 +0x1c4 fp=0xc000047b68 sp=0xc000047b18 pc=0x7ff686827c44
net/http.serverHandler.ServeHTTP({0x7ff687f9eeb0?}, {0x7ff687fa3020?, 0xc000692000?}, 0x1?)
        net/http/server.go:3301 +0x8e fp=0xc000047b98 sp=0xc000047b68 pc=0x7ff6868456ce
net/http.(*conn).serve(0xc000532090, {0x7ff687fa5978, 0xc000512390})
        net/http/server.go:2102 +0x625 fp=0xc000047fb8 sp=0xc000047b98 pc=0x7ff686824245
net/http.(*Server).Serve.gowrap3()
        net/http/server.go:3454 +0x28 fp=0xc000047fe0 sp=0xc000047fb8 pc=0x7ff686829b08
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000047fe8 sp=0xc000047fe0 pc=0x7ff68650db21
created by net/http.(*Server).Serve in goroutine 1
        net/http/server.go:3454 +0x485

goroutine 895 gp=0xc000507dc0 m=nil [IO wait]:
runtime.gopark(0x0?, 0xc00051e2a0?, 0x48?, 0xe3?, 0xc00051e34c?)
        runtime/proc.go:435 +0xce fp=0xc001725d58 sp=0xc001725d38 pc=0x7ff68650598e
runtime.netpollblock(0x2ac?, 0x864a0406?, 0xf6?)
        runtime/netpoll.go:575 +0xf7 fp=0xc001725d90 sp=0xc001725d58 pc=0x7ff6864cbdf7
internal/poll.runtime_pollWait(0x1dff7d11858, 0x72)
        runtime/netpoll.go:351 +0x85 fp=0xc001725db0 sp=0xc001725d90 pc=0x7ff686504b25
internal/poll.(*pollDesc).wait(0x0?, 0xc000532100?, 0x0)
        internal/poll/fd_poll_runtime.go:84 +0x27 fp=0xc001725dd8 sp=0xc001725db0 pc=0x7ff68659c0c7
internal/poll.execIO(0xc00051e2a0, 0x7ff687dd8778)
        internal/poll/fd_windows.go:177 +0x105 fp=0xc001725e50 sp=0xc001725dd8 pc=0x7ff68659d525
internal/poll.(*FD).Read(0xc00051e288, {0xc0003d20a1, 0x1, 0x1})
        internal/poll/fd_windows.go:438 +0x29b fp=0xc001725ef0 sp=0xc001725e50 pc=0x7ff68659e1fb
net.(*netFD).Read(0xc00051e288, {0xc0003d20a1?, 0x0?, 0x0?})
        net/fd_posix.go:55 +0x25 fp=0xc001725f38 sp=0xc001725ef0 pc=0x7ff686611945
net.(*conn).Read(0xc000602000, {0xc0003d20a1?, 0x0?, 0x0?})
        net/net.go:194 +0x45 fp=0xc001725f80 sp=0xc001725f38 pc=0x7ff686621065
net/http.(*connReader).backgroundRead(0xc0003d2090)
        net/http/server.go:690 +0x37 fp=0xc001725fc8 sp=0xc001725f80 pc=0x7ff68681e117
net/http.(*connReader).startBackgroundRead.gowrap2()
        net/http/server.go:686 +0x25 fp=0xc001725fe0 sp=0xc001725fc8 pc=0x7ff68681e045
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc001725fe8 sp=0xc001725fe0 pc=0x7ff68650db21
created by net/http.(*connReader).startBackgroundRead in goroutine 15
        net/http/server.go:686 +0xb6

goroutine 1032 gp=0xc00098c380 m=nil [chan receive]:
runtime.gopark(0x30?, 0x7ff687c8ea20?, 0x1?, 0x9d?, 0xc00041bb20?)
        runtime/proc.go:435 +0xce fp=0xc00041bad8 sp=0xc00041bab8 pc=0x7ff68650598e
runtime.chanrecv(0xc000690230, 0x0, 0x1)
        runtime/chan.go:664 +0x445 fp=0xc00041bb50 sp=0xc00041bad8 pc=0x7ff6864a2d45
runtime.chanrecv1(0x7ff687d96d37?, 0x2c?)
        runtime/chan.go:506 +0x12 fp=0xc00041bb78 sp=0xc00041bb50 pc=0x7ff6864a28d2
github.com/ollama/ollama/runner/ollamarunner.(*Server).computeBatch(0xc00051c000, {0x1, {0x7ff687fb3400, 0xc009726000}, {0x7ff687fc3298, 0xc000589d70}, {0xc0009b6008, 0x200, 0x25f}, {{0x7ff687fc3298, ...}, ...}, ...})
        github.com/ollama/ollama/runner/ollamarunner/runner.go:645 +0x185 fp=0xc00041bef0 sp=0xc00041bb78 pc=0x7ff686ba0f05
github.com/ollama/ollama/runner/ollamarunner.(*Server).run.gowrap1()
        github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x58 fp=0xc00041bfe0 sp=0xc00041bef0 pc=0x7ff686b9f2d8
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00041bfe8 sp=0xc00041bfe0 pc=0x7ff68650db21
created by github.com/ollama/ollama/runner/ollamarunner.(*Server).run in goroutine 50
        github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x2cd
rax     0x0
rbx     0x3f79aff628
rcx     0x700000006
rdx     0x0
rdi     0xe06d7363
rsi     0x1
rbp     0x4
rsp     0x3f79aff500
r8      0x1
r9      0xe06d7363
r10     0x0
r11     0x1dfb20d4980
r12     0x0
r13     0x1e0458e5e90
r14     0x8000
r15     0x0
rip     0x7ffe3ffd79da
rflags  0x206
cs      0x33
fs      0x53
gs      0x2b
time=2026-04-30T11:45:18.251+08:00 level=ERROR source=server.go:1624 msg="post predict" error="Post \"http://127.0.0.1:3238/completion\": read tcp 127.0.0.1:3245->127.0.0.1:3238: wsarecv: An existing connection was forcibly closed by the remote host."
[GIN] 2026/04/30 - 11:45:18 | 500 |   16.2620155s |       127.0.0.1 | POST     "/v1/chat/completions"
time=2026-04-30T11:45:19.380+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3297"
time=2026-04-30T11:45:20.377+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3316"
time=2026-04-30T11:45:21.071+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3331"
time=2026-04-30T11:45:21.775+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3346"
time=2026-04-30T11:45:22.457+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3357"
time=2026-04-30T11:45:23.173+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3377"
time=2026-04-30T11:45:23.828+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3388"
time=2026-04-30T11:45:24.528+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3400"
time=2026-04-30T11:45:25.125+08:00 level=INFO source=runner.go:464 msg="failure during GPU discovery" OLLAMA_LIBRARY_PATH="[C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\lib\\ollama C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\lib\\ollama\\vulkan]" extra_envs=map[] error="failed to finish discovery before timeout"
time=2026-04-30T11:45:25.125+08:00 level=WARN source=runner.go:356 msg="unable to refresh free memory, using old values"
time=2026-04-30T11:45:25.127+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3408"
time=2026-04-30T11:45:25.772+08:00 level=INFO source=cpu_windows.go:148 msg=packages count=1
time=2026-04-30T11:45:25.772+08:00 level=INFO source=cpu_windows.go:164 msg="efficiency cores detected" maxEfficiencyClass=1
time=2026-04-30T11:45:25.772+08:00 level=INFO source=cpu_windows.go:195 msg="" package=0 cores=10 efficiency=4 threads=16
time=2026-04-30T11:45:26.023+08:00 level=INFO source=model.go:97 msg="gemma4: token IDs" image=255999 image_end=258882 audio=256000 audio_end=258883
time=2026-04-30T11:45:26.024+08:00 level=INFO source=server.go:259 msg="enabling flash attention"
time=2026-04-30T11:45:26.025+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --model D:\\AI_MODELS\\models\\blobs\\sha256-7121486771cbfe218851513210c40b35dbdee93ab1ef43fe36283c883980f0df --port 3427"
time=2026-04-30T11:45:26.053+08:00 level=INFO source=sched.go:484 msg="system memory" total="47.7 GiB" free="31.5 GiB" free_swap="46.7 GiB"
time=2026-04-30T11:45:26.053+08:00 level=INFO source=sched.go:491 msg="gpu memory" id=8680a156-0800-0000-0300-000000000000 library=Vulkan available="4.9 GiB" free="5.4 GiB" minimum="457.0 MiB" overhead="0 B"
time=2026-04-30T11:45:26.053+08:00 level=INFO source=server.go:771 msg="loading model" "model layers"=31 requested=-1
time=2026-04-30T11:45:26.103+08:00 level=INFO source=runner.go:1417 msg="starting ollama engine"
time=2026-04-30T11:45:26.111+08:00 level=INFO source=runner.go:1452 msg="Server listening on 127.0.0.1:3427"
time=2026-04-30T11:45:26.121+08:00 level=INFO source=runner.go:1290 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:31[ID:8680a156-0800-0000-0300-000000000000 Layers:31(0..30)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
time=2026-04-30T11:45:26.174+08:00 level=INFO source=ggml.go:136 msg="" architecture=gemma4 file_type=Q4_K_M name="" description="" num_tensors=1014 num_key_values=52
load_backend: loaded CPU backend from C:\Users\Administrator\AppData\Local\Programs\Ollama\lib\ollama\ggml-cpu-alderlake.dll
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = Intel(R) Arc(TM) A750 Graphics (Intel Corporation) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: none
load_backend: loaded Vulkan backend from C:\Users\Administrator\AppData\Local\Programs\Ollama\lib\ollama\vulkan\ggml-vulkan.dll
time=2026-04-30T11:45:26.288+08:00 level=INFO source=ggml.go:104 msg=system CPU.0.SSE3=1 CPU.0.SSSE3=1 CPU.0.AVX=1 CPU.0.AVX_VNNI=1 CPU.0.AVX2=1 CPU.0.F16C=1 CPU.0.FMA=1 CPU.0.BMI2=1 CPU.0.LLAMAFILE=1 CPU.1.LLAMAFILE=1 compiler=cgo(clang)
ggml_backend_vk_get_device_memory called: uuid 8680a156-0800-0000-0300-000000000000
ggml_backend_vk_get_device_memory called: luid 0x000000000000f6b6
ggml_dxgi_pdh_init called
DXGI + PDH Initialized. Getting GPU free memory info
[DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000000F6B6, Dedicated: 7.91 GB, Shared: 23.87 GB
[DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000001C965, Dedicated: 7.91 GB, Shared: 23.87 GB
[DXGI] Adapter Description: Microsoft Basic Render Driver, LUID: 0x000000000000FAF4, Dedicated: 0.00 GB, Shared: 23.87 GB
Discrete GPU (Intel(R) Arc(TM) A750 Graphics) with LUID 0x000000000000f6b6 detected. Dedicated Total: 8491368448.00 bytes (7.91 GB), Dedicated Usage: 2732904448.00 bytes (2.55 GB)
ggml_backend_vk_get_device_memory utilizing DXGI + PDH memory reporting free: 5758464000 total: 8491368448
time=2026-04-30T11:45:26.533+08:00 level=INFO source=model.go:97 msg="gemma4: token IDs" image=255999 image_end=258882 audio=256000 audio_end=258883
time=2026-04-30T11:45:26.547+08:00 level=INFO source=model.go:138 msg="vision: decode" elapsed=2.1271ms bounds=(0,0)-(2048,2048)
time=2026-04-30T11:45:26.648+08:00 level=INFO source=model.go:145 msg="vision: preprocess" elapsed=101.1927ms size="[768 768]"
time=2026-04-30T11:45:26.648+08:00 level=INFO source=model.go:148 msg="vision: pixelValues" shape="[768 768 3]" dim0=768 dim1=768 dim2=3
time=2026-04-30T11:45:26.648+08:00 level=INFO source=model.go:152 msg="vision: patches" patchesX=48 patchesY=48 total=2304 patchSize=16
time=2026-04-30T11:45:26.649+08:00 level=INFO source=model.go:156 msg="vision: encoded" elapsed=104.8169ms shape="[2816 256]"
time=2026-04-30T11:45:26.675+08:00 level=INFO source=runner.go:1290 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:8[ID:8680a156-0800-0000-0300-000000000000 Layers:8(22..29)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
ggml_backend_vk_get_device_memory called: uuid 8680a156-0800-0000-0300-000000000000
ggml_backend_vk_get_device_memory called: luid 0x000000000000f6b6
ggml_dxgi_pdh_init called
DXGI + PDH Initialized. Getting GPU free memory info
[DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000000F6B6, Dedicated: 7.91 GB, Shared: 23.87 GB
[DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000001C965, Dedicated: 7.91 GB, Shared: 23.87 GB
[DXGI] Adapter Description: Microsoft Basic Render Driver, LUID: 0x000000000000FAF4, Dedicated: 0.00 GB, Shared: 23.87 GB
Discrete GPU (Intel(R) Arc(TM) A750 Graphics) with LUID 0x000000000000f6b6 detected. Dedicated Total: 8491368448.00 bytes (7.91 GB), Dedicated Usage: 2732904448.00 bytes (2.55 GB)
ggml_backend_vk_get_device_memory utilizing DXGI + PDH memory reporting free: 5758464000 total: 8491368448
time=2026-04-30T11:45:26.809+08:00 level=INFO source=model.go:97 msg="gemma4: token IDs" image=255999 image_end=258882 audio=256000 audio_end=258883
time=2026-04-30T11:45:26.823+08:00 level=INFO source=model.go:138 msg="vision: decode" elapsed=1.0667ms bounds=(0,0)-(2048,2048)
time=2026-04-30T11:45:26.920+08:00 level=INFO source=model.go:145 msg="vision: preprocess" elapsed=97.1047ms size="[768 768]"
time=2026-04-30T11:45:26.920+08:00 level=INFO source=model.go:148 msg="vision: pixelValues" shape="[768 768 3]" dim0=768 dim1=768 dim2=3
time=2026-04-30T11:45:26.920+08:00 level=INFO source=model.go:152 msg="vision: patches" patchesX=48 patchesY=48 total=2304 patchSize=16
time=2026-04-30T11:45:26.921+08:00 level=INFO source=model.go:156 msg="vision: encoded" elapsed=99.0688ms shape="[2816 256]"
time=2026-04-30T11:45:26.947+08:00 level=INFO source=runner.go:1290 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:8[ID:8680a156-0800-0000-0300-000000000000 Layers:8(22..29)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
ggml_backend_vk_get_device_memory called: uuid 8680a156-0800-0000-0300-000000000000
ggml_backend_vk_get_device_memory called: luid 0x000000000000f6b6
ggml_dxgi_pdh_init called
DXGI + PDH Initialized. Getting GPU free memory info
[DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000000F6B6, Dedicated: 7.91 GB, Shared: 23.87 GB
[DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000001C965, Dedicated: 7.91 GB, Shared: 23.87 GB
[DXGI] Adapter Description: Microsoft Basic Render Driver, LUID: 0x000000000000FAF4, Dedicated: 0.00 GB, Shared: 23.87 GB
Discrete GPU (Intel(R) Arc(TM) A750 Graphics) with LUID 0x000000000000f6b6 detected. Dedicated Total: 8491368448.00 bytes (7.91 GB), Dedicated Usage: 2732904448.00 bytes (2.55 GB)
ggml_backend_vk_get_device_memory utilizing DXGI + PDH memory reporting free: 5758464000 total: 8491368448
time=2026-04-30T11:45:27.352+08:00 level=INFO source=model.go:97 msg="gemma4: token IDs" image=255999 image_end=258882 audio=256000 audio_end=258883
time=2026-04-30T11:45:27.361+08:00 level=INFO source=model.go:138 msg="vision: decode" elapsed=1.049ms bounds=(0,0)-(2048,2048)
time=2026-04-30T11:45:27.461+08:00 level=INFO source=model.go:145 msg="vision: preprocess" elapsed=99.9253ms size="[768 768]"
time=2026-04-30T11:45:27.463+08:00 level=INFO source=model.go:148 msg="vision: pixelValues" shape="[768 768 3]" dim0=768 dim1=768 dim2=3
time=2026-04-30T11:45:27.463+08:00 level=INFO source=model.go:152 msg="vision: patches" patchesX=48 patchesY=48 total=2304 patchSize=16
time=2026-04-30T11:45:27.464+08:00 level=INFO source=model.go:156 msg="vision: encoded" elapsed=103.7039ms shape="[2816 256]"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=runner.go:1290 msg=load request="{Operation:commit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:8[ID:8680a156-0800-0000-0300-000000000000 Layers:8(22..29)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=ggml.go:482 msg="offloading 8 repeating layers to GPU"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=ggml.go:486 msg="offloading output layer to CPU"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=ggml.go:494 msg="offloaded 8/31 layers to GPU"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=device.go:240 msg="model weights" device=Vulkan0 size="4.1 GiB"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=device.go:245 msg="model weights" device=CPU size="13.2 GiB"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=device.go:251 msg="kv cache" device=Vulkan0 size="224.0 MiB"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=device.go:256 msg="kv cache" device=CPU size="656.0 MiB"
time=2026-04-30T11:45:27.691+08:00 level=INFO source=device.go:262 msg="compute graph" device=Vulkan0 size="352.3 MiB"
time=2026-04-30T11:45:27.691+08:00 level=INFO source=device.go:267 msg="compute graph" device=CPU size="55.0 MiB"
time=2026-04-30T11:45:27.692+08:00 level=INFO source=device.go:272 msg="total memory" size="18.6 GiB"
time=2026-04-30T11:45:27.692+08:00 level=INFO source=sched.go:561 msg="loaded runners" count=1
time=2026-04-30T11:45:27.693+08:00 level=INFO source=server.go:1364 msg="waiting for llama runner to start responding"
time=2026-04-30T11:45:27.695+08:00 level=INFO source=server.go:1398 msg="waiting for server to become available" status="llm server loading model"
time=2026-04-30T11:45:31.144+08:00 level=INFO source=server.go:1402 msg="llama runner started in 5.09 seconds"
Exception 0xe06d7363 0x19930520 0x4b92aff500 0x7ffe3ffd79da
PC=0x7ffe3ffd79da
signal arrived during external code execution

runtime.cgocall(0x7ff687680aa0, 0xc000f0daa0)
        runtime/cgocall.go:167 +0x3e fp=0xc000f0da78 sp=0xc000f0da10 pc=0x7ff68650243e
github.com/ollama/ollama/ml/backend/ggml._Cfunc_ggml_backend_sched_graph_compute_async(0x17ac69a07c0, 0x17f45d3f600)
        _cgo_gotypes.go:963 +0x50 fp=0xc000f0daa0 sp=0xc000f0da78 pc=0x7ff686a42370
github.com/ollama/ollama/ml/backend/ggml.(*Context).ComputeWithNotify.func2(...)
        github.com/ollama/ollama/ml/backend/ggml/ggml.go:825
github.com/ollama/ollama/ml/backend/ggml.(*Context).ComputeWithNotify(0xc00972c840, 0xc0002db800?, {0xc000591950, 0x1, 0x2?})
        github.com/ollama/ollama/ml/backend/ggml/ggml.go:825 +0x1b5 fp=0xc000f0db78 sp=0xc000f0daa0 pc=0x7ff686a520b5
github.com/ollama/ollama/runner/ollamarunner.(*Server).computeBatch(0xc0004c6000, {0x0, {0x7ff687fb3400, 0xc00972c840}, {0x7ff687fc3298, 0xc0097e2de0}, {0xc001d0c608, 0x200, 0x25f}, {{0x7ff687fc3298, ...}, ...}, ...})
        github.com/ollama/ollama/runner/ollamarunner/runner.go:716 +0x862 fp=0xc000f0def0 sp=0xc000f0db78 pc=0x7ff686ba15e2
github.com/ollama/ollama/runner/ollamarunner.(*Server).run.gowrap1()
        github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x58 fp=0xc000f0dfe0 sp=0xc000f0def0 pc=0x7ff686b9f2d8
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000f0dfe8 sp=0xc000f0dfe0 pc=0x7ff68650db21
created by github.com/ollama/ollama/runner/ollamarunner.(*Server).run in goroutine 37
        github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x2cd

goroutine 1 gp=0xc0000021c0 m=nil [IO wait, locked to thread]:
runtime.gopark(0x7ff68650f320?, 0x7ff688b5c280?, 0x20?, 0x80?, 0xc0004c80cc?)
        runtime/proc.go:435 +0xce fp=0xc001685630 sp=0xc001685610 pc=0x7ff68650598e
runtime.netpollblock(0x26c?, 0x864a0406?, 0xf6?)
        runtime/netpoll.go:575 +0xf7 fp=0xc001685668 sp=0xc001685630 pc=0x7ff6864cbdf7
internal/poll.runtime_pollWait(0x17a78d7e8f0, 0x72)
        runtime/netpoll.go:351 +0x85 fp=0xc001685688 sp=0xc001685668 pc=0x7ff686504b25
internal/poll.(*pollDesc).wait(0x7ff68659aad3?, 0x0?, 0x0)
        internal/poll/fd_poll_runtime.go:84 +0x27 fp=0xc0016856b0 sp=0xc001685688 pc=0x7ff68659c0c7
internal/poll.execIO(0xc0004c8020, 0xc0003a7758)
        internal/poll/fd_windows.go:177 +0x105 fp=0xc001685728 sp=0xc0016856b0 pc=0x7ff68659d525
internal/poll.(*FD).acceptOne(0xc0004c8008, 0x2b8, {0xc0004d40f0?, 0xc0003a77b8?, 0x7ff6865a51e5?}, 0xc0003a77ec?)
        internal/poll/fd_windows.go:946 +0x65 fp=0xc001685788 sp=0xc001685728 pc=0x7ff6865a1aa5
internal/poll.(*FD).Accept(0xc0004c8008, 0xc001685938)
        internal/poll/fd_windows.go:980 +0x1b6 fp=0xc001685840 sp=0xc001685788 pc=0x7ff6865a1dd6
net.(*netFD).accept(0xc0004c8008)
        net/fd_windows.go:182 +0x4b fp=0xc001685958 sp=0xc001685840 pc=0x7ff686613a6b
net.(*TCPListener).accept(0xc0001c4080)
        net/tcpsock_posix.go:159 +0x1b fp=0xc0016859a8 sp=0xc001685958 pc=0x7ff68662a01b
net.(*TCPListener).Accept(0xc0001c4080)
        net/tcpsock.go:380 +0x30 fp=0xc0016859d8 sp=0xc0016859a8 pc=0x7ff686628dd0
net/http.(*onceCloseListener).Accept(0xc0004d2090?)
        <autogenerated>:1 +0x24 fp=0xc0016859f0 sp=0xc0016859d8 pc=0x7ff686851e44
net/http.(*Server).Serve(0xc000690400, {0x7ff687fa2e70, 0xc0001c4080})
        net/http/server.go:3424 +0x30c fp=0xc001685b20 sp=0xc0016859f0 pc=0x7ff68682970c
github.com/ollama/ollama/runner/ollamarunner.Execute({0xc00004c0b0, 0x4, 0x5})
        github.com/ollama/ollama/runner/ollamarunner/runner.go:1453 +0x94e fp=0xc001685cf0 sp=0xc001685b20 pc=0x7ff686ba866e
github.com/ollama/ollama/runner.Execute({0xc00004c090?, 0x0?, 0x0?})
        github.com/ollama/ollama/runner/runner.go:18 +0x10e fp=0xc001685d30 sp=0xc001685cf0 pc=0x7ff686c6d0ae
github.com/ollama/ollama/cmd.NewCLI.func3(0xc00014d800?, {0x7ff687d47a56?, 0x4?, 0x7ff687d47a5a?})
        github.com/ollama/ollama/cmd/cmd.go:2327 +0x45 fp=0xc001685d58 sp=0xc001685d30 pc=0x7ff6874dbaa5
github.com/spf13/cobra.(*Command).execute(0xc00047bb08, {0xc000696d70, 0x5, 0x5})
        github.com/spf13/cobra@v1.7.0/command.go:940 +0x85c fp=0xc001685e78 sp=0xc001685d58 pc=0x7ff68668ec3c
github.com/spf13/cobra.(*Command).ExecuteC(0xc000420908)
        github.com/spf13/cobra@v1.7.0/command.go:1068 +0x3a5 fp=0xc001685f30 sp=0xc001685e78 pc=0x7ff68668f485
github.com/spf13/cobra.(*Command).Execute(...)
        github.com/spf13/cobra@v1.7.0/command.go:992
github.com/spf13/cobra.(*Command).ExecuteContext(...)
        github.com/spf13/cobra@v1.7.0/command.go:985
main.main()
        github.com/ollama/ollama/main.go:12 +0x4d fp=0xc001685f50 sp=0xc001685f30 pc=0x7ff6874dd42d
runtime.main()
        runtime/proc.go:283 +0x27d fp=0xc001685fe0 sp=0xc001685f50 pc=0x7ff6864d4ddd
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc001685fe8 sp=0xc001685fe0 pc=0x7ff68650db21

goroutine 2 gp=0xc0000028c0 m=nil [force gc (idle)]:
runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000079fa8 sp=0xc000079f88 pc=0x7ff68650598e
runtime.goparkunlock(...)
        runtime/proc.go:441
runtime.forcegchelper()
        runtime/proc.go:348 +0xb8 fp=0xc000079fe0 sp=0xc000079fa8 pc=0x7ff6864d50f8
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000079fe8 sp=0xc000079fe0 pc=0x7ff68650db21
created by runtime.init.7 in goroutine 1
        runtime/proc.go:336 +0x1a

goroutine 3 gp=0xc000002c40 m=nil [GC sweep wait]:
runtime.gopark(0x1?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00007bf80 sp=0xc00007bf60 pc=0x7ff68650598e
runtime.goparkunlock(...)
        runtime/proc.go:441
runtime.bgsweep(0xc000088000)
        runtime/mgcsweep.go:316 +0xdf fp=0xc00007bfc8 sp=0xc00007bf80 pc=0x7ff6864bdebf
runtime.gcenable.gowrap1()
        runtime/mgc.go:204 +0x25 fp=0xc00007bfe0 sp=0xc00007bfc8 pc=0x7ff6864b2285
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00007bfe8 sp=0xc00007bfe0 pc=0x7ff68650db21
created by runtime.gcenable in goroutine 1
        runtime/mgc.go:204 +0x66

goroutine 4 gp=0xc000002e00 m=nil [GC scavenge wait]:
runtime.gopark(0x277c34?, 0x7267d5?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00008ff78 sp=0xc00008ff58 pc=0x7ff68650598e
runtime.goparkunlock(...)
        runtime/proc.go:441
runtime.(*scavengerState).park(0x7ff688b8a740)
        runtime/mgcscavenge.go:425 +0x49 fp=0xc00008ffa8 sp=0xc00008ff78 pc=0x7ff6864bb909
runtime.bgscavenge(0xc000088000)
        runtime/mgcscavenge.go:658 +0x59 fp=0xc00008ffc8 sp=0xc00008ffa8 pc=0x7ff6864bbe99
runtime.gcenable.gowrap2()
        runtime/mgc.go:205 +0x25 fp=0xc00008ffe0 sp=0xc00008ffc8 pc=0x7ff6864b2225
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00008ffe8 sp=0xc00008ffe0 pc=0x7ff68650db21
created by runtime.gcenable in goroutine 1
        runtime/mgc.go:205 +0xa5

goroutine 5 gp=0xc000003340 m=nil [finalizer wait]:
runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000091e30 sp=0xc000091e10 pc=0x7ff68650598e
runtime.runfinq()
        runtime/mfinal.go:196 +0x107 fp=0xc000091fe0 sp=0xc000091e30 pc=0x7ff6864b1207
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000091fe8 sp=0xc000091fe0 pc=0x7ff68650db21
created by runtime.createfing in goroutine 1
        runtime/mfinal.go:166 +0x3d

goroutine 6 gp=0xc000003dc0 m=nil [chan receive]:
runtime.gopark(0xc0000e1900?, 0xc0097e2e28?, 0x60?, 0xdf?, 0x7ff6865fc688?)
        runtime/proc.go:435 +0xce fp=0xc00007df18 sp=0xc00007def8 pc=0x7ff68650598e
runtime.chanrecv(0xc00003a5b0, 0x0, 0x1)
        runtime/chan.go:664 +0x445 fp=0xc00007df90 sp=0xc00007df18 pc=0x7ff6864a2d45
runtime.chanrecv1(0x7ff6864d4f40?, 0xc00007df76?)
        runtime/chan.go:506 +0x12 fp=0xc00007dfb8 sp=0xc00007df90 pc=0x7ff6864a28d2
runtime.unique_runtime_registerUniqueMapCleanup.func2(...)
        runtime/mgc.go:1796
runtime.unique_runtime_registerUniqueMapCleanup.gowrap1()
        runtime/mgc.go:1799 +0x2f fp=0xc00007dfe0 sp=0xc00007dfb8 pc=0x7ff6864b54af
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00007dfe8 sp=0xc00007dfe0 pc=0x7ff68650db21
created by unique.runtime_registerUniqueMapCleanup in goroutine 1
        runtime/mgc.go:1794 +0x85

goroutine 18 gp=0xc0001f8000 m=nil [GC worker (idle)]:
runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00008bf38 sp=0xc00008bf18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc00008bfc8 sp=0xc00008bf38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc00008bfe0 sp=0xc00008bfc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00008bfe8 sp=0xc00008bfe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 7 gp=0xc0004021c0 m=nil [GC worker (idle)]:
runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000409f38 sp=0xc000409f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000409fc8 sp=0xc000409f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000409fe0 sp=0xc000409fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000409fe8 sp=0xc000409fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 8 gp=0xc000402380 m=nil [GC worker (idle)]:
runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00040bf38 sp=0xc00040bf18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc00040bfc8 sp=0xc00040bf38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc00040bfe0 sp=0xc00040bfc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00040bfe8 sp=0xc00040bfe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 9 gp=0xc000402540 m=nil [GC worker (idle)]:
runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000405f38 sp=0xc000405f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000405fc8 sp=0xc000405f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000405fe0 sp=0xc000405fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000405fe8 sp=0xc000405fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 10 gp=0xc000402700 m=nil [GC worker (idle)]:
runtime.gopark(0x57dd78c1be4?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000407f38 sp=0xc000407f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000407fc8 sp=0xc000407f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000407fe0 sp=0xc000407fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000407fe8 sp=0xc000407fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 11 gp=0xc0004028c0 m=nil [GC worker (idle)]:
runtime.gopark(0x57ce6b85278?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000411f38 sp=0xc000411f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000411fc8 sp=0xc000411f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000411fe0 sp=0xc000411fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000411fe8 sp=0xc000411fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 34 gp=0xc000484000 m=nil [GC worker (idle)]:
runtime.gopark(0x7ff688bdd9c0?, 0x1?, 0x40?, 0x1b?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00040df38 sp=0xc00040df18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc00040dfc8 sp=0xc00040df38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc00040dfe0 sp=0xc00040dfc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00040dfe8 sp=0xc00040dfe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 19 gp=0xc0001f81c0 m=nil [GC worker (idle)]:
runtime.gopark(0x57dd78c1be4?, 0x3?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00008df38 sp=0xc00008df18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc00008dfc8 sp=0xc00008df38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc00008dfe0 sp=0xc00008dfc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00008dfe8 sp=0xc00008dfe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 12 gp=0xc000402a80 m=nil [GC worker (idle)]:
runtime.gopark(0x57dd78c1be4?, 0x1?, 0xcc?, 0x2?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000413f38 sp=0xc000413f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000413fc8 sp=0xc000413f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000413fe0 sp=0xc000413fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000413fe8 sp=0xc000413fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 13 gp=0xc000402c40 m=nil [GC worker (idle)]:
runtime.gopark(0x57ce6b85278?, 0x1?, 0x88?, 0x2?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000419f38 sp=0xc000419f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000419fc8 sp=0xc000419f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000419fe0 sp=0xc000419fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000419fe8 sp=0xc000419fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 35 gp=0xc0004841c0 m=nil [GC worker (idle)]:
runtime.gopark(0x57ce6b85278?, 0x1?, 0xa8?, 0x1c?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00040ff38 sp=0xc00040ff18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc00040ffc8 sp=0xc00040ff38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc00040ffe0 sp=0xc00040ffc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00040ffe8 sp=0xc00040ffe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 20 gp=0xc0001f8380 m=nil [GC worker (idle)]:
runtime.gopark(0x57ce6b85278?, 0x3?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000415f38 sp=0xc000415f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000415fc8 sp=0xc000415f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000415fe0 sp=0xc000415fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000415fe8 sp=0xc000415fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 14 gp=0xc000402e00 m=nil [GC worker (idle)]:
runtime.gopark(0x57dd78c1be4?, 0x1?, 0x3c?, 0x31?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00041bf38 sp=0xc00041bf18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc00041bfc8 sp=0xc00041bf38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc00041bfe0 sp=0xc00041bfc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00041bfe8 sp=0xc00041bfe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 36 gp=0xc000484380 m=nil [GC worker (idle)]:
runtime.gopark(0x57dd8158f00?, 0x1?, 0x0?, 0x1a?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00048bf38 sp=0xc00048bf18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc00048bfc8 sp=0xc00048bf38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc00048bfe0 sp=0xc00048bfc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00048bfe8 sp=0xc00048bfe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 21 gp=0xc0001f8540 m=nil [GC worker (idle)]:
runtime.gopark(0x57dd77cbf78?, 0x3?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000417f38 sp=0xc000417f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000417fc8 sp=0xc000417f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000417fe0 sp=0xc000417fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000417fe8 sp=0xc000417fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 15 gp=0xc000402fc0 m=nil [GC worker (idle)]:
runtime.gopark(0x7ff688bdd9c0?, 0x1?, 0x28?, 0x2e?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000487f38 sp=0xc000487f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000487fc8 sp=0xc000487f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000487fe0 sp=0xc000487fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000487fe8 sp=0xc000487fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 37 gp=0xc0001f88c0 m=nil [chan receive]:
runtime.gopark(0x30?, 0xffffffffffffffff?, 0x3?, 0x0?, 0xc001fa7798?)
        runtime/proc.go:435 +0xce fp=0xc001fa7750 sp=0xc001fa7730 pc=0x7ff68650598e
runtime.chanrecv(0xc00003a230, 0x0, 0x1)
        runtime/chan.go:664 +0x445 fp=0xc001fa77c8 sp=0xc001fa7750 pc=0x7ff6864a2d45
runtime.chanrecv1(0x7ff687d92ab6?, 0x29?)
        runtime/chan.go:506 +0x12 fp=0xc001fa77f0 sp=0xc001fa77c8 pc=0x7ff6864a28d2
github.com/ollama/ollama/runner/ollamarunner.(*Server).forwardBatch(_, {0x1, {0x7ff687fb3400, 0xc0096f8340}, {0x7ff687fc3298, 0xc0097c64f8}, {0xc00096b308, 0x200, 0x25f}, {{0x7ff687fc3298, ...}, ...}, ...})
        github.com/ollama/ollama/runner/ollamarunner/runner.go:476 +0xfa fp=0xc001fa7b58 sp=0xc001fa77f0 pc=0x7ff686b9f3fa
github.com/ollama/ollama/runner/ollamarunner.(*Server).run(0xc0004c6000, {0x7ff687fa59b0, 0xc0004c4000})
        github.com/ollama/ollama/runner/ollamarunner/runner.go:453 +0x18c fp=0xc001fa7fb8 sp=0xc001fa7b58 pc=0x7ff686b9f0ac
github.com/ollama/ollama/runner/ollamarunner.Execute.gowrap1()
        github.com/ollama/ollama/runner/ollamarunner/runner.go:1430 +0x28 fp=0xc001fa7fe0 sp=0xc001fa7fb8 pc=0x7ff686ba88e8
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc001fa7fe8 sp=0xc001fa7fe0 pc=0x7ff68650db21
created by github.com/ollama/ollama/runner/ollamarunner.Execute in goroutine 1
        github.com/ollama/ollama/runner/ollamarunner/runner.go:1430 +0x4c9

goroutine 38 gp=0xc0001f8a80 m=nil [select]:
runtime.gopark(0xc000047a08?, 0x2?, 0xc0?, 0x77?, 0xc00004786c?)
        runtime/proc.go:435 +0xce fp=0xc001687698 sp=0xc001687678 pc=0x7ff68650598e
runtime.selectgo(0xc001687a08, 0xc000047868, 0xe3c?, 0x0, 0x1?, 0x1)
        runtime/select.go:351 +0x837 fp=0xc0016877d0 sp=0xc001687698 pc=0x7ff6864e6437
github.com/ollama/ollama/runner/ollamarunner.(*Server).completion(0xc0004c6000, {0x7ff687fa3020, 0xc0002d01c0}, 0xc0003de3c0)
        github.com/ollama/ollama/runner/ollamarunner/runner.go:956 +0xc4e fp=0xc001687ac0 sp=0xc0016877d0 pc=0x7ff686ba388e
github.com/ollama/ollama/runner/ollamarunner.(*Server).completion-fm({0x7ff687fa3020?, 0xc0002d01c0?}, 0xc000047b40?)
        <autogenerated>:1 +0x36 fp=0xc001687af0 sp=0xc001687ac0 pc=0x7ff686ba8dd6
net/http.HandlerFunc.ServeHTTP(0xc0004ca000?, {0x7ff687fa3020?, 0xc0002d01c0?}, 0xc000047b60?)
        net/http/server.go:2294 +0x29 fp=0xc001687b18 sp=0xc001687af0 pc=0x7ff686825d49
net/http.(*ServeMux).ServeHTTP(0x7ff6864ab785?, {0x7ff687fa3020, 0xc0002d01c0}, 0xc0003de3c0)
        net/http/server.go:2822 +0x1c4 fp=0xc001687b68 sp=0xc001687b18 pc=0x7ff686827c44
net/http.serverHandler.ServeHTTP({0x7ff687f9eeb0?}, {0x7ff687fa3020?, 0xc0002d01c0?}, 0x1?)
        net/http/server.go:3301 +0x8e fp=0xc001687b98 sp=0xc001687b68 pc=0x7ff6868456ce
net/http.(*conn).serve(0xc0004d2090, {0x7ff687fa5978, 0xc0004c0360})
        net/http/server.go:2102 +0x625 fp=0xc001687fb8 sp=0xc001687b98 pc=0x7ff686824245
net/http.(*Server).Serve.gowrap3()
        net/http/server.go:3454 +0x28 fp=0xc001687fe0 sp=0xc001687fb8 pc=0x7ff686829b08
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc001687fe8 sp=0xc001687fe0 pc=0x7ff68650db21
created by net/http.(*Server).Serve in goroutine 1
        net/http/server.go:3454 +0x485

goroutine 1142 gp=0xc000485880 m=nil [IO wait]:
runtime.gopark(0x0?, 0xc0004c82a0?, 0x48?, 0x83?, 0xc0004c834c?)
        runtime/proc.go:435 +0xce fp=0xc001097d58 sp=0xc001097d38 pc=0x7ff68650598e
runtime.netpollblock(0x214?, 0x864a0406?, 0xf6?)
        runtime/netpoll.go:575 +0xf7 fp=0xc001097d90 sp=0xc001097d58 pc=0x7ff6864cbdf7
internal/poll.runtime_pollWait(0x17a78d7e7d8, 0x72)
        runtime/netpoll.go:351 +0x85 fp=0xc001097db0 sp=0xc001097d90 pc=0x7ff686504b25
internal/poll.(*pollDesc).wait(0x214?, 0x72?, 0x0)
        internal/poll/fd_poll_runtime.go:84 +0x27 fp=0xc001097dd8 sp=0xc001097db0 pc=0x7ff68659c0c7
internal/poll.execIO(0xc0004c82a0, 0x7ff687dd8778)
        internal/poll/fd_windows.go:177 +0x105 fp=0xc001097e50 sp=0xc001097dd8 pc=0x7ff68659d525
internal/poll.(*FD).Read(0xc0004c8288, {0xc000188041, 0x1, 0x1})
        internal/poll/fd_windows.go:438 +0x29b fp=0xc001097ef0 sp=0xc001097e50 pc=0x7ff68659e1fb
net.(*netFD).Read(0xc0004c8288, {0xc000188041?, 0xc00068e758?, 0xc001097f70?})
        net/fd_posix.go:55 +0x25 fp=0xc001097f38 sp=0xc001097ef0 pc=0x7ff686611945
net.(*conn).Read(0xc00020c090, {0xc000188041?, 0x7ff687facb50?, 0xc00029e320?})
        net/net.go:194 +0x45 fp=0xc001097f80 sp=0xc001097f38 pc=0x7ff686621065
net/http.(*connReader).backgroundRead(0xc000188030)
        net/http/server.go:690 +0x37 fp=0xc001097fc8 sp=0xc001097f80 pc=0x7ff68681e117
net/http.(*connReader).startBackgroundRead.gowrap2()
        net/http/server.go:686 +0x25 fp=0xc001097fe0 sp=0xc001097fc8 pc=0x7ff68681e045
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc001097fe8 sp=0xc001097fe0 pc=0x7ff68650db21
created by net/http.(*connReader).startBackgroundRead in goroutine 38
        net/http/server.go:686 +0xb6

goroutine 1172 gp=0xc00058ea80 m=nil [chan receive]:
runtime.gopark(0x30?, 0x247ff01?, 0x60?, 0x4?, 0xc0001adb20?)
        runtime/proc.go:435 +0xce fp=0xc0001adad8 sp=0xc0001adab8 pc=0x7ff68650598e
runtime.chanrecv(0xc0003201c0, 0x0, 0x1)
        runtime/chan.go:664 +0x445 fp=0xc0001adb50 sp=0xc0001adad8 pc=0x7ff6864a2d45
runtime.chanrecv1(0x7ff687d96d37?, 0x2c?)
        runtime/chan.go:506 +0x12 fp=0xc0001adb78 sp=0xc0001adb50 pc=0x7ff6864a28d2
github.com/ollama/ollama/runner/ollamarunner.(*Server).computeBatch(0xc0004c6000, {0x1, {0x7ff687fb3400, 0xc0096f8340}, {0x7ff687fc3298, 0xc0097c64f8}, {0xc00096b308, 0x200, 0x25f}, {{0x7ff687fc3298, ...}, ...}, ...})
        github.com/ollama/ollama/runner/ollamarunner/runner.go:645 +0x185 fp=0xc0001adef0 sp=0xc0001adb78 pc=0x7ff686ba0f05
github.com/ollama/ollama/runner/ollamarunner.(*Server).run.gowrap1()
        github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x58 fp=0xc0001adfe0 sp=0xc0001adef0 pc=0x7ff686b9f2d8
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc0001adfe8 sp=0xc0001adfe0 pc=0x7ff68650db21
created by github.com/ollama/ollama/runner/ollamarunner.(*Server).run in goroutine 37
        github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x2cd
rax     0x0
rbx     0x4b92aff488
rcx     0x700000006
rdx     0x0
rdi     0xe06d7363
rsi     0x1
rbp     0x4
rsp     0x4b92aff360
r8      0x1
r9      0xe06d7363
r10     0x0
r11     0x17a732c2100
r12     0x0
r13     0x17ac69a07c0
r14     0x1000000
r15     0x0
rip     0x7ffe3ffd79da
rflags  0x206
cs      0x33
fs      0x53
gs      0x2b
time=2026-04-30T11:45:34.878+08:00 level=ERROR source=server.go:1624 msg="post predict" error="Post \"http://127.0.0.1:3427/completion\": read tcp 127.0.0.1:3434->127.0.0.1:3427: wsarecv: An existing connection was forcibly closed by the remote host."
[GIN] 2026/04/30 - 11:45:34 | 500 |   15.7829065s |       127.0.0.1 | POST     "/v1/chat/completions"

RAW_BUFFERClick to expand / collapse

What is the issue?

Description

This is a follow-up to #15888. After further investigation, the issue is not just "internal label leakage" but a complete crash of the Vulkan backend when running gemma4 26b MoE with image input on Intel ARC A750.

Symptoms

When the crash occurs: Ollama runner process terminates with a C++ exception during GPU computation

Observable behavior: The runner repeatedly crashes and restarts, appearing as "反复载入再踹掉" (repeated model loading/unloading) on every API request

Final result: API returns HTTP 500 error

Crash Signature

Exception 0xe06d7363 0x19930520 0x4b92aff500 0x7ffe3ffd79da PC=0x7ffe3ffd79da signal arrived during external code execution

Exception 0xe06d7363 is a C++ exception (thrown from native code, not an access violation).

Stack Trace (relevant frames)

github.com/ollama/ollama/ml/backend/ggml._Cfunc_ggml_backend_sched_graph_compute_async _cgo_gotypes.go:963 +0x50 github.com/ollama/ollama/ml/backend/ggml.(.ComputeWithNotify ggml.go:825 +0x1b5 github.com/ollama/ollama/runner/ollamarunner.(.computeBatch runner.go:716 +0x862 github.com/ollama/ollama/runner/ollamarunner.(.run.gowrap1() runner.go:459 +0x58

The crash occurs inside ggml_backend_sched_graph_compute_async (Vulkan backend).

Log Excerpt

time=... level=INFO source=ggml.go:136 msg="" architecture=gemma4 file_type=Q4_K_M ggml_vulkan: Found 1 Vulkan devices: ggml_vulkan: 0 = Intel(R) Arc(TM) A750 Graphics (Intel Corporation) | uma: 0 | fp16: 1 | bf16: 0 load_backend: loaded Vulkan backend from ...\ggml-vulkan.dll ... time=... level=INFO source=runner.go:1290 msg="load request" GPULayers=8 ... Exception 0xe06d7363 ... PC=0x7ffe3ffd79da time=... level=ERROR source=server.go:1624 msg="post predict" error="... wsarecv: An existing connection was forcibly closed by the remote host." [GIN] ... | 500 | ... | POST "/v1/chat/completions"

Steps to Reproduce

Hardware: Intel ARC A750

Ollama version: v0.20 (Windows)

Model: gemma4:26b-A4B (MoE variant)

Ensure Vulkan backend is active (it appears to be the default on this hardware in v0.20): ollama serve

Send an image via API: curl http://localhost:11434/api/generate -d "{"model":"gemma4:26b-A4B","prompt":"Describe this image","images":["base64_image_data"]}"

Expected Behavior

Model generates a response without crashing.

Actual Behavior

Runner process crashes with C++ exception

Ollama service restarts the runner repeatedly

Each request fails with HTTP 500

Workaround

Disabling Vulkan backend (forcing CPU) resolves the issue:

set OLLAMA_GPU_OVERHEAD=0 set OLLAMA_NUM_GPU=0 ollama serve

Or rename the Vulkan DLL to prevent loading.

Environment

OS: Windows

GPU: Intel ARC A750

CPU: Intel

Ollama version: v0.22

Backend: Vulkan (ggml-vulkan.dll)

Additional Context

This crash only occurs with gemma4 26b MoE + image input + Vulkan backend

The same crash pattern was observed as "repeated model loading/unloading" from the user's perspective

The issue may be related to the earlier label leakage report (#15888) – possibly the same root cause manifesting as either label leakage or full crash depending on timing/memory state

Related Issues

#15888 – Internal label leakage with gemma4 + Vulkan on ARC A750

Relevant log output

0ac
github.com/ollama/ollama/runner/ollamarunner.Execute.gowrap1()
        github.com/ollama/ollama/runner/ollamarunner/runner.go:1430 +0x28 fp=0xc00086ffe0 sp=0xc00086ffb8 pc=0x7ff686ba88e8
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00086ffe8 sp=0xc00086ffe0 pc=0x7ff68650db21
created by github.com/ollama/ollama/runner/ollamarunner.Execute in goroutine 1
        github.com/ollama/ollama/runner/ollamarunner/runner.go:1430 +0x4c9

goroutine 15 gp=0xc0004168c0 m=nil [select]:
runtime.gopark(0xc000047a08?, 0x2?, 0xc0?, 0x77?, 0xc00004786c?)
        runtime/proc.go:435 +0xce fp=0xc000047698 sp=0xc000047678 pc=0x7ff68650598e
runtime.selectgo(0xc000047a08, 0xc000047868, 0xe3c?, 0x0, 0x1?, 0x1)
        runtime/select.go:351 +0x837 fp=0xc0000477d0 sp=0xc000047698 pc=0x7ff6864e6437
github.com/ollama/ollama/runner/ollamarunner.(*Server).completion(0xc00051c000, {0x7ff687fa3020, 0xc000692000}, 0xc000616000)
        github.com/ollama/ollama/runner/ollamarunner/runner.go:956 +0xc4e fp=0xc000047ac0 sp=0xc0000477d0 pc=0x7ff686ba388e
github.com/ollama/ollama/runner/ollamarunner.(*Server).completion-fm({0x7ff687fa3020?, 0xc000692000?}, 0xc000047b40?)
        <autogenerated>:1 +0x36 fp=0xc000047af0 sp=0xc000047ac0 pc=0x7ff686ba8dd6
net/http.HandlerFunc.ServeHTTP(0xc000520000?, {0x7ff687fa3020?, 0xc000692000?}, 0xc000047b60?)
        net/http/server.go:2294 +0x29 fp=0xc000047b18 sp=0xc000047af0 pc=0x7ff686825d49
net/http.(*ServeMux).ServeHTTP(0x7ff6864ab785?, {0x7ff687fa3020, 0xc000692000}, 0xc000616000)
        net/http/server.go:2822 +0x1c4 fp=0xc000047b68 sp=0xc000047b18 pc=0x7ff686827c44
net/http.serverHandler.ServeHTTP({0x7ff687f9eeb0?}, {0x7ff687fa3020?, 0xc000692000?}, 0x1?)
        net/http/server.go:3301 +0x8e fp=0xc000047b98 sp=0xc000047b68 pc=0x7ff6868456ce
net/http.(*conn).serve(0xc000532090, {0x7ff687fa5978, 0xc000512390})
        net/http/server.go:2102 +0x625 fp=0xc000047fb8 sp=0xc000047b98 pc=0x7ff686824245
net/http.(*Server).Serve.gowrap3()
        net/http/server.go:3454 +0x28 fp=0xc000047fe0 sp=0xc000047fb8 pc=0x7ff686829b08
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000047fe8 sp=0xc000047fe0 pc=0x7ff68650db21
created by net/http.(*Server).Serve in goroutine 1
        net/http/server.go:3454 +0x485

goroutine 895 gp=0xc000507dc0 m=nil [IO wait]:
runtime.gopark(0x0?, 0xc00051e2a0?, 0x48?, 0xe3?, 0xc00051e34c?)
        runtime/proc.go:435 +0xce fp=0xc001725d58 sp=0xc001725d38 pc=0x7ff68650598e
runtime.netpollblock(0x2ac?, 0x864a0406?, 0xf6?)
        runtime/netpoll.go:575 +0xf7 fp=0xc001725d90 sp=0xc001725d58 pc=0x7ff6864cbdf7
internal/poll.runtime_pollWait(0x1dff7d11858, 0x72)
        runtime/netpoll.go:351 +0x85 fp=0xc001725db0 sp=0xc001725d90 pc=0x7ff686504b25
internal/poll.(*pollDesc).wait(0x0?, 0xc000532100?, 0x0)
        internal/poll/fd_poll_runtime.go:84 +0x27 fp=0xc001725dd8 sp=0xc001725db0 pc=0x7ff68659c0c7
internal/poll.execIO(0xc00051e2a0, 0x7ff687dd8778)
        internal/poll/fd_windows.go:177 +0x105 fp=0xc001725e50 sp=0xc001725dd8 pc=0x7ff68659d525
internal/poll.(*FD).Read(0xc00051e288, {0xc0003d20a1, 0x1, 0x1})
        internal/poll/fd_windows.go:438 +0x29b fp=0xc001725ef0 sp=0xc001725e50 pc=0x7ff68659e1fb
net.(*netFD).Read(0xc00051e288, {0xc0003d20a1?, 0x0?, 0x0?})
        net/fd_posix.go:55 +0x25 fp=0xc001725f38 sp=0xc001725ef0 pc=0x7ff686611945
net.(*conn).Read(0xc000602000, {0xc0003d20a1?, 0x0?, 0x0?})
        net/net.go:194 +0x45 fp=0xc001725f80 sp=0xc001725f38 pc=0x7ff686621065
net/http.(*connReader).backgroundRead(0xc0003d2090)
        net/http/server.go:690 +0x37 fp=0xc001725fc8 sp=0xc001725f80 pc=0x7ff68681e117
net/http.(*connReader).startBackgroundRead.gowrap2()
        net/http/server.go:686 +0x25 fp=0xc001725fe0 sp=0xc001725fc8 pc=0x7ff68681e045
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc001725fe8 sp=0xc001725fe0 pc=0x7ff68650db21
created by net/http.(*connReader).startBackgroundRead in goroutine 15
        net/http/server.go:686 +0xb6

goroutine 1032 gp=0xc00098c380 m=nil [chan receive]:
runtime.gopark(0x30?, 0x7ff687c8ea20?, 0x1?, 0x9d?, 0xc00041bb20?)
        runtime/proc.go:435 +0xce fp=0xc00041bad8 sp=0xc00041bab8 pc=0x7ff68650598e
runtime.chanrecv(0xc000690230, 0x0, 0x1)
        runtime/chan.go:664 +0x445 fp=0xc00041bb50 sp=0xc00041bad8 pc=0x7ff6864a2d45
runtime.chanrecv1(0x7ff687d96d37?, 0x2c?)
        runtime/chan.go:506 +0x12 fp=0xc00041bb78 sp=0xc00041bb50 pc=0x7ff6864a28d2
github.com/ollama/ollama/runner/ollamarunner.(*Server).computeBatch(0xc00051c000, {0x1, {0x7ff687fb3400, 0xc009726000}, {0x7ff687fc3298, 0xc000589d70}, {0xc0009b6008, 0x200, 0x25f}, {{0x7ff687fc3298, ...}, ...}, ...})
        github.com/ollama/ollama/runner/ollamarunner/runner.go:645 +0x185 fp=0xc00041bef0 sp=0xc00041bb78 pc=0x7ff686ba0f05
github.com/ollama/ollama/runner/ollamarunner.(*Server).run.gowrap1()
        github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x58 fp=0xc00041bfe0 sp=0xc00041bef0 pc=0x7ff686b9f2d8
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00041bfe8 sp=0xc00041bfe0 pc=0x7ff68650db21
created by github.com/ollama/ollama/runner/ollamarunner.(*Server).run in goroutine 50
        github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x2cd
rax     0x0
rbx     0x3f79aff628
rcx     0x700000006
rdx     0x0
rdi     0xe06d7363
rsi     0x1
rbp     0x4
rsp     0x3f79aff500
r8      0x1
r9      0xe06d7363
r10     0x0
r11     0x1dfb20d4980
r12     0x0
r13     0x1e0458e5e90
r14     0x8000
r15     0x0
rip     0x7ffe3ffd79da
rflags  0x206
cs      0x33
fs      0x53
gs      0x2b
time=2026-04-30T11:45:18.251+08:00 level=ERROR source=server.go:1624 msg="post predict" error="Post \"http://127.0.0.1:3238/completion\": read tcp 127.0.0.1:3245->127.0.0.1:3238: wsarecv: An existing connection was forcibly closed by the remote host."
[GIN] 2026/04/30 - 11:45:18 | 500 |   16.2620155s |       127.0.0.1 | POST     "/v1/chat/completions"
time=2026-04-30T11:45:19.380+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3297"
time=2026-04-30T11:45:20.377+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3316"
time=2026-04-30T11:45:21.071+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3331"
time=2026-04-30T11:45:21.775+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3346"
time=2026-04-30T11:45:22.457+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3357"
time=2026-04-30T11:45:23.173+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3377"
time=2026-04-30T11:45:23.828+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3388"
time=2026-04-30T11:45:24.528+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3400"
time=2026-04-30T11:45:25.125+08:00 level=INFO source=runner.go:464 msg="failure during GPU discovery" OLLAMA_LIBRARY_PATH="[C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\lib\\ollama C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\lib\\ollama\\vulkan]" extra_envs=map[] error="failed to finish discovery before timeout"
time=2026-04-30T11:45:25.125+08:00 level=WARN source=runner.go:356 msg="unable to refresh free memory, using old values"
time=2026-04-30T11:45:25.127+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 3408"
time=2026-04-30T11:45:25.772+08:00 level=INFO source=cpu_windows.go:148 msg=packages count=1
time=2026-04-30T11:45:25.772+08:00 level=INFO source=cpu_windows.go:164 msg="efficiency cores detected" maxEfficiencyClass=1
time=2026-04-30T11:45:25.772+08:00 level=INFO source=cpu_windows.go:195 msg="" package=0 cores=10 efficiency=4 threads=16
time=2026-04-30T11:45:26.023+08:00 level=INFO source=model.go:97 msg="gemma4: token IDs" image=255999 image_end=258882 audio=256000 audio_end=258883
time=2026-04-30T11:45:26.024+08:00 level=INFO source=server.go:259 msg="enabling flash attention"
time=2026-04-30T11:45:26.025+08:00 level=INFO source=server.go:444 msg="starting runner" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --model D:\\AI_MODELS\\models\\blobs\\sha256-7121486771cbfe218851513210c40b35dbdee93ab1ef43fe36283c883980f0df --port 3427"
time=2026-04-30T11:45:26.053+08:00 level=INFO source=sched.go:484 msg="system memory" total="47.7 GiB" free="31.5 GiB" free_swap="46.7 GiB"
time=2026-04-30T11:45:26.053+08:00 level=INFO source=sched.go:491 msg="gpu memory" id=8680a156-0800-0000-0300-000000000000 library=Vulkan available="4.9 GiB" free="5.4 GiB" minimum="457.0 MiB" overhead="0 B"
time=2026-04-30T11:45:26.053+08:00 level=INFO source=server.go:771 msg="loading model" "model layers"=31 requested=-1
time=2026-04-30T11:45:26.103+08:00 level=INFO source=runner.go:1417 msg="starting ollama engine"
time=2026-04-30T11:45:26.111+08:00 level=INFO source=runner.go:1452 msg="Server listening on 127.0.0.1:3427"
time=2026-04-30T11:45:26.121+08:00 level=INFO source=runner.go:1290 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:31[ID:8680a156-0800-0000-0300-000000000000 Layers:31(0..30)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
time=2026-04-30T11:45:26.174+08:00 level=INFO source=ggml.go:136 msg="" architecture=gemma4 file_type=Q4_K_M name="" description="" num_tensors=1014 num_key_values=52
load_backend: loaded CPU backend from C:\Users\Administrator\AppData\Local\Programs\Ollama\lib\ollama\ggml-cpu-alderlake.dll
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = Intel(R) Arc(TM) A750 Graphics (Intel Corporation) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: none
load_backend: loaded Vulkan backend from C:\Users\Administrator\AppData\Local\Programs\Ollama\lib\ollama\vulkan\ggml-vulkan.dll
time=2026-04-30T11:45:26.288+08:00 level=INFO source=ggml.go:104 msg=system CPU.0.SSE3=1 CPU.0.SSSE3=1 CPU.0.AVX=1 CPU.0.AVX_VNNI=1 CPU.0.AVX2=1 CPU.0.F16C=1 CPU.0.FMA=1 CPU.0.BMI2=1 CPU.0.LLAMAFILE=1 CPU.1.LLAMAFILE=1 compiler=cgo(clang)
ggml_backend_vk_get_device_memory called: uuid 8680a156-0800-0000-0300-000000000000
ggml_backend_vk_get_device_memory called: luid 0x000000000000f6b6
ggml_dxgi_pdh_init called
DXGI + PDH Initialized. Getting GPU free memory info
[DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000000F6B6, Dedicated: 7.91 GB, Shared: 23.87 GB
[DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000001C965, Dedicated: 7.91 GB, Shared: 23.87 GB
[DXGI] Adapter Description: Microsoft Basic Render Driver, LUID: 0x000000000000FAF4, Dedicated: 0.00 GB, Shared: 23.87 GB
Discrete GPU (Intel(R) Arc(TM) A750 Graphics) with LUID 0x000000000000f6b6 detected. Dedicated Total: 8491368448.00 bytes (7.91 GB), Dedicated Usage: 2732904448.00 bytes (2.55 GB)
ggml_backend_vk_get_device_memory utilizing DXGI + PDH memory reporting free: 5758464000 total: 8491368448
time=2026-04-30T11:45:26.533+08:00 level=INFO source=model.go:97 msg="gemma4: token IDs" image=255999 image_end=258882 audio=256000 audio_end=258883
time=2026-04-30T11:45:26.547+08:00 level=INFO source=model.go:138 msg="vision: decode" elapsed=2.1271ms bounds=(0,0)-(2048,2048)
time=2026-04-30T11:45:26.648+08:00 level=INFO source=model.go:145 msg="vision: preprocess" elapsed=101.1927ms size="[768 768]"
time=2026-04-30T11:45:26.648+08:00 level=INFO source=model.go:148 msg="vision: pixelValues" shape="[768 768 3]" dim0=768 dim1=768 dim2=3
time=2026-04-30T11:45:26.648+08:00 level=INFO source=model.go:152 msg="vision: patches" patchesX=48 patchesY=48 total=2304 patchSize=16
time=2026-04-30T11:45:26.649+08:00 level=INFO source=model.go:156 msg="vision: encoded" elapsed=104.8169ms shape="[2816 256]"
time=2026-04-30T11:45:26.675+08:00 level=INFO source=runner.go:1290 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:8[ID:8680a156-0800-0000-0300-000000000000 Layers:8(22..29)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
ggml_backend_vk_get_device_memory called: uuid 8680a156-0800-0000-0300-000000000000
ggml_backend_vk_get_device_memory called: luid 0x000000000000f6b6
ggml_dxgi_pdh_init called
DXGI + PDH Initialized. Getting GPU free memory info
[DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000000F6B6, Dedicated: 7.91 GB, Shared: 23.87 GB
[DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000001C965, Dedicated: 7.91 GB, Shared: 23.87 GB
[DXGI] Adapter Description: Microsoft Basic Render Driver, LUID: 0x000000000000FAF4, Dedicated: 0.00 GB, Shared: 23.87 GB
Discrete GPU (Intel(R) Arc(TM) A750 Graphics) with LUID 0x000000000000f6b6 detected. Dedicated Total: 8491368448.00 bytes (7.91 GB), Dedicated Usage: 2732904448.00 bytes (2.55 GB)
ggml_backend_vk_get_device_memory utilizing DXGI + PDH memory reporting free: 5758464000 total: 8491368448
time=2026-04-30T11:45:26.809+08:00 level=INFO source=model.go:97 msg="gemma4: token IDs" image=255999 image_end=258882 audio=256000 audio_end=258883
time=2026-04-30T11:45:26.823+08:00 level=INFO source=model.go:138 msg="vision: decode" elapsed=1.0667ms bounds=(0,0)-(2048,2048)
time=2026-04-30T11:45:26.920+08:00 level=INFO source=model.go:145 msg="vision: preprocess" elapsed=97.1047ms size="[768 768]"
time=2026-04-30T11:45:26.920+08:00 level=INFO source=model.go:148 msg="vision: pixelValues" shape="[768 768 3]" dim0=768 dim1=768 dim2=3
time=2026-04-30T11:45:26.920+08:00 level=INFO source=model.go:152 msg="vision: patches" patchesX=48 patchesY=48 total=2304 patchSize=16
time=2026-04-30T11:45:26.921+08:00 level=INFO source=model.go:156 msg="vision: encoded" elapsed=99.0688ms shape="[2816 256]"
time=2026-04-30T11:45:26.947+08:00 level=INFO source=runner.go:1290 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:8[ID:8680a156-0800-0000-0300-000000000000 Layers:8(22..29)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
ggml_backend_vk_get_device_memory called: uuid 8680a156-0800-0000-0300-000000000000
ggml_backend_vk_get_device_memory called: luid 0x000000000000f6b6
ggml_dxgi_pdh_init called
DXGI + PDH Initialized. Getting GPU free memory info
[DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000000F6B6, Dedicated: 7.91 GB, Shared: 23.87 GB
[DXGI] Adapter Description: Intel(R) Arc(TM) A750 Graphics, LUID: 0x000000000001C965, Dedicated: 7.91 GB, Shared: 23.87 GB
[DXGI] Adapter Description: Microsoft Basic Render Driver, LUID: 0x000000000000FAF4, Dedicated: 0.00 GB, Shared: 23.87 GB
Discrete GPU (Intel(R) Arc(TM) A750 Graphics) with LUID 0x000000000000f6b6 detected. Dedicated Total: 8491368448.00 bytes (7.91 GB), Dedicated Usage: 2732904448.00 bytes (2.55 GB)
ggml_backend_vk_get_device_memory utilizing DXGI + PDH memory reporting free: 5758464000 total: 8491368448
time=2026-04-30T11:45:27.352+08:00 level=INFO source=model.go:97 msg="gemma4: token IDs" image=255999 image_end=258882 audio=256000 audio_end=258883
time=2026-04-30T11:45:27.361+08:00 level=INFO source=model.go:138 msg="vision: decode" elapsed=1.049ms bounds=(0,0)-(2048,2048)
time=2026-04-30T11:45:27.461+08:00 level=INFO source=model.go:145 msg="vision: preprocess" elapsed=99.9253ms size="[768 768]"
time=2026-04-30T11:45:27.463+08:00 level=INFO source=model.go:148 msg="vision: pixelValues" shape="[768 768 3]" dim0=768 dim1=768 dim2=3
time=2026-04-30T11:45:27.463+08:00 level=INFO source=model.go:152 msg="vision: patches" patchesX=48 patchesY=48 total=2304 patchSize=16
time=2026-04-30T11:45:27.464+08:00 level=INFO source=model.go:156 msg="vision: encoded" elapsed=103.7039ms shape="[2816 256]"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=runner.go:1290 msg=load request="{Operation:commit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:6 GPULayers:8[ID:8680a156-0800-0000-0300-000000000000 Layers:8(22..29)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=ggml.go:482 msg="offloading 8 repeating layers to GPU"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=ggml.go:486 msg="offloading output layer to CPU"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=ggml.go:494 msg="offloaded 8/31 layers to GPU"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=device.go:240 msg="model weights" device=Vulkan0 size="4.1 GiB"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=device.go:245 msg="model weights" device=CPU size="13.2 GiB"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=device.go:251 msg="kv cache" device=Vulkan0 size="224.0 MiB"
time=2026-04-30T11:45:27.690+08:00 level=INFO source=device.go:256 msg="kv cache" device=CPU size="656.0 MiB"
time=2026-04-30T11:45:27.691+08:00 level=INFO source=device.go:262 msg="compute graph" device=Vulkan0 size="352.3 MiB"
time=2026-04-30T11:45:27.691+08:00 level=INFO source=device.go:267 msg="compute graph" device=CPU size="55.0 MiB"
time=2026-04-30T11:45:27.692+08:00 level=INFO source=device.go:272 msg="total memory" size="18.6 GiB"
time=2026-04-30T11:45:27.692+08:00 level=INFO source=sched.go:561 msg="loaded runners" count=1
time=2026-04-30T11:45:27.693+08:00 level=INFO source=server.go:1364 msg="waiting for llama runner to start responding"
time=2026-04-30T11:45:27.695+08:00 level=INFO source=server.go:1398 msg="waiting for server to become available" status="llm server loading model"
time=2026-04-30T11:45:31.144+08:00 level=INFO source=server.go:1402 msg="llama runner started in 5.09 seconds"
Exception 0xe06d7363 0x19930520 0x4b92aff500 0x7ffe3ffd79da
PC=0x7ffe3ffd79da
signal arrived during external code execution

runtime.cgocall(0x7ff687680aa0, 0xc000f0daa0)
        runtime/cgocall.go:167 +0x3e fp=0xc000f0da78 sp=0xc000f0da10 pc=0x7ff68650243e
github.com/ollama/ollama/ml/backend/ggml._Cfunc_ggml_backend_sched_graph_compute_async(0x17ac69a07c0, 0x17f45d3f600)
        _cgo_gotypes.go:963 +0x50 fp=0xc000f0daa0 sp=0xc000f0da78 pc=0x7ff686a42370
github.com/ollama/ollama/ml/backend/ggml.(*Context).ComputeWithNotify.func2(...)
        github.com/ollama/ollama/ml/backend/ggml/ggml.go:825
github.com/ollama/ollama/ml/backend/ggml.(*Context).ComputeWithNotify(0xc00972c840, 0xc0002db800?, {0xc000591950, 0x1, 0x2?})
        github.com/ollama/ollama/ml/backend/ggml/ggml.go:825 +0x1b5 fp=0xc000f0db78 sp=0xc000f0daa0 pc=0x7ff686a520b5
github.com/ollama/ollama/runner/ollamarunner.(*Server).computeBatch(0xc0004c6000, {0x0, {0x7ff687fb3400, 0xc00972c840}, {0x7ff687fc3298, 0xc0097e2de0}, {0xc001d0c608, 0x200, 0x25f}, {{0x7ff687fc3298, ...}, ...}, ...})
        github.com/ollama/ollama/runner/ollamarunner/runner.go:716 +0x862 fp=0xc000f0def0 sp=0xc000f0db78 pc=0x7ff686ba15e2
github.com/ollama/ollama/runner/ollamarunner.(*Server).run.gowrap1()
        github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x58 fp=0xc000f0dfe0 sp=0xc000f0def0 pc=0x7ff686b9f2d8
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000f0dfe8 sp=0xc000f0dfe0 pc=0x7ff68650db21
created by github.com/ollama/ollama/runner/ollamarunner.(*Server).run in goroutine 37
        github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x2cd

goroutine 1 gp=0xc0000021c0 m=nil [IO wait, locked to thread]:
runtime.gopark(0x7ff68650f320?, 0x7ff688b5c280?, 0x20?, 0x80?, 0xc0004c80cc?)
        runtime/proc.go:435 +0xce fp=0xc001685630 sp=0xc001685610 pc=0x7ff68650598e
runtime.netpollblock(0x26c?, 0x864a0406?, 0xf6?)
        runtime/netpoll.go:575 +0xf7 fp=0xc001685668 sp=0xc001685630 pc=0x7ff6864cbdf7
internal/poll.runtime_pollWait(0x17a78d7e8f0, 0x72)
        runtime/netpoll.go:351 +0x85 fp=0xc001685688 sp=0xc001685668 pc=0x7ff686504b25
internal/poll.(*pollDesc).wait(0x7ff68659aad3?, 0x0?, 0x0)
        internal/poll/fd_poll_runtime.go:84 +0x27 fp=0xc0016856b0 sp=0xc001685688 pc=0x7ff68659c0c7
internal/poll.execIO(0xc0004c8020, 0xc0003a7758)
        internal/poll/fd_windows.go:177 +0x105 fp=0xc001685728 sp=0xc0016856b0 pc=0x7ff68659d525
internal/poll.(*FD).acceptOne(0xc0004c8008, 0x2b8, {0xc0004d40f0?, 0xc0003a77b8?, 0x7ff6865a51e5?}, 0xc0003a77ec?)
        internal/poll/fd_windows.go:946 +0x65 fp=0xc001685788 sp=0xc001685728 pc=0x7ff6865a1aa5
internal/poll.(*FD).Accept(0xc0004c8008, 0xc001685938)
        internal/poll/fd_windows.go:980 +0x1b6 fp=0xc001685840 sp=0xc001685788 pc=0x7ff6865a1dd6
net.(*netFD).accept(0xc0004c8008)
        net/fd_windows.go:182 +0x4b fp=0xc001685958 sp=0xc001685840 pc=0x7ff686613a6b
net.(*TCPListener).accept(0xc0001c4080)
        net/tcpsock_posix.go:159 +0x1b fp=0xc0016859a8 sp=0xc001685958 pc=0x7ff68662a01b
net.(*TCPListener).Accept(0xc0001c4080)
        net/tcpsock.go:380 +0x30 fp=0xc0016859d8 sp=0xc0016859a8 pc=0x7ff686628dd0
net/http.(*onceCloseListener).Accept(0xc0004d2090?)
        <autogenerated>:1 +0x24 fp=0xc0016859f0 sp=0xc0016859d8 pc=0x7ff686851e44
net/http.(*Server).Serve(0xc000690400, {0x7ff687fa2e70, 0xc0001c4080})
        net/http/server.go:3424 +0x30c fp=0xc001685b20 sp=0xc0016859f0 pc=0x7ff68682970c
github.com/ollama/ollama/runner/ollamarunner.Execute({0xc00004c0b0, 0x4, 0x5})
        github.com/ollama/ollama/runner/ollamarunner/runner.go:1453 +0x94e fp=0xc001685cf0 sp=0xc001685b20 pc=0x7ff686ba866e
github.com/ollama/ollama/runner.Execute({0xc00004c090?, 0x0?, 0x0?})
        github.com/ollama/ollama/runner/runner.go:18 +0x10e fp=0xc001685d30 sp=0xc001685cf0 pc=0x7ff686c6d0ae
github.com/ollama/ollama/cmd.NewCLI.func3(0xc00014d800?, {0x7ff687d47a56?, 0x4?, 0x7ff687d47a5a?})
        github.com/ollama/ollama/cmd/cmd.go:2327 +0x45 fp=0xc001685d58 sp=0xc001685d30 pc=0x7ff6874dbaa5
github.com/spf13/cobra.(*Command).execute(0xc00047bb08, {0xc000696d70, 0x5, 0x5})
        github.com/spf13/[email protected]/command.go:940 +0x85c fp=0xc001685e78 sp=0xc001685d58 pc=0x7ff68668ec3c
github.com/spf13/cobra.(*Command).ExecuteC(0xc000420908)
        github.com/spf13/[email protected]/command.go:1068 +0x3a5 fp=0xc001685f30 sp=0xc001685e78 pc=0x7ff68668f485
github.com/spf13/cobra.(*Command).Execute(...)
        github.com/spf13/[email protected]/command.go:992
github.com/spf13/cobra.(*Command).ExecuteContext(...)
        github.com/spf13/[email protected]/command.go:985
main.main()
        github.com/ollama/ollama/main.go:12 +0x4d fp=0xc001685f50 sp=0xc001685f30 pc=0x7ff6874dd42d
runtime.main()
        runtime/proc.go:283 +0x27d fp=0xc001685fe0 sp=0xc001685f50 pc=0x7ff6864d4ddd
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc001685fe8 sp=0xc001685fe0 pc=0x7ff68650db21

goroutine 2 gp=0xc0000028c0 m=nil [force gc (idle)]:
runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000079fa8 sp=0xc000079f88 pc=0x7ff68650598e
runtime.goparkunlock(...)
        runtime/proc.go:441
runtime.forcegchelper()
        runtime/proc.go:348 +0xb8 fp=0xc000079fe0 sp=0xc000079fa8 pc=0x7ff6864d50f8
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000079fe8 sp=0xc000079fe0 pc=0x7ff68650db21
created by runtime.init.7 in goroutine 1
        runtime/proc.go:336 +0x1a

goroutine 3 gp=0xc000002c40 m=nil [GC sweep wait]:
runtime.gopark(0x1?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00007bf80 sp=0xc00007bf60 pc=0x7ff68650598e
runtime.goparkunlock(...)
        runtime/proc.go:441
runtime.bgsweep(0xc000088000)
        runtime/mgcsweep.go:316 +0xdf fp=0xc00007bfc8 sp=0xc00007bf80 pc=0x7ff6864bdebf
runtime.gcenable.gowrap1()
        runtime/mgc.go:204 +0x25 fp=0xc00007bfe0 sp=0xc00007bfc8 pc=0x7ff6864b2285
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00007bfe8 sp=0xc00007bfe0 pc=0x7ff68650db21
created by runtime.gcenable in goroutine 1
        runtime/mgc.go:204 +0x66

goroutine 4 gp=0xc000002e00 m=nil [GC scavenge wait]:
runtime.gopark(0x277c34?, 0x7267d5?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00008ff78 sp=0xc00008ff58 pc=0x7ff68650598e
runtime.goparkunlock(...)
        runtime/proc.go:441
runtime.(*scavengerState).park(0x7ff688b8a740)
        runtime/mgcscavenge.go:425 +0x49 fp=0xc00008ffa8 sp=0xc00008ff78 pc=0x7ff6864bb909
runtime.bgscavenge(0xc000088000)
        runtime/mgcscavenge.go:658 +0x59 fp=0xc00008ffc8 sp=0xc00008ffa8 pc=0x7ff6864bbe99
runtime.gcenable.gowrap2()
        runtime/mgc.go:205 +0x25 fp=0xc00008ffe0 sp=0xc00008ffc8 pc=0x7ff6864b2225
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00008ffe8 sp=0xc00008ffe0 pc=0x7ff68650db21
created by runtime.gcenable in goroutine 1
        runtime/mgc.go:205 +0xa5

goroutine 5 gp=0xc000003340 m=nil [finalizer wait]:
runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000091e30 sp=0xc000091e10 pc=0x7ff68650598e
runtime.runfinq()
        runtime/mfinal.go:196 +0x107 fp=0xc000091fe0 sp=0xc000091e30 pc=0x7ff6864b1207
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000091fe8 sp=0xc000091fe0 pc=0x7ff68650db21
created by runtime.createfing in goroutine 1
        runtime/mfinal.go:166 +0x3d

goroutine 6 gp=0xc000003dc0 m=nil [chan receive]:
runtime.gopark(0xc0000e1900?, 0xc0097e2e28?, 0x60?, 0xdf?, 0x7ff6865fc688?)
        runtime/proc.go:435 +0xce fp=0xc00007df18 sp=0xc00007def8 pc=0x7ff68650598e
runtime.chanrecv(0xc00003a5b0, 0x0, 0x1)
        runtime/chan.go:664 +0x445 fp=0xc00007df90 sp=0xc00007df18 pc=0x7ff6864a2d45
runtime.chanrecv1(0x7ff6864d4f40?, 0xc00007df76?)
        runtime/chan.go:506 +0x12 fp=0xc00007dfb8 sp=0xc00007df90 pc=0x7ff6864a28d2
runtime.unique_runtime_registerUniqueMapCleanup.func2(...)
        runtime/mgc.go:1796
runtime.unique_runtime_registerUniqueMapCleanup.gowrap1()
        runtime/mgc.go:1799 +0x2f fp=0xc00007dfe0 sp=0xc00007dfb8 pc=0x7ff6864b54af
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00007dfe8 sp=0xc00007dfe0 pc=0x7ff68650db21
created by unique.runtime_registerUniqueMapCleanup in goroutine 1
        runtime/mgc.go:1794 +0x85

goroutine 18 gp=0xc0001f8000 m=nil [GC worker (idle)]:
runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00008bf38 sp=0xc00008bf18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc00008bfc8 sp=0xc00008bf38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc00008bfe0 sp=0xc00008bfc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00008bfe8 sp=0xc00008bfe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 7 gp=0xc0004021c0 m=nil [GC worker (idle)]:
runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000409f38 sp=0xc000409f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000409fc8 sp=0xc000409f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000409fe0 sp=0xc000409fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000409fe8 sp=0xc000409fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 8 gp=0xc000402380 m=nil [GC worker (idle)]:
runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00040bf38 sp=0xc00040bf18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc00040bfc8 sp=0xc00040bf38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc00040bfe0 sp=0xc00040bfc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00040bfe8 sp=0xc00040bfe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 9 gp=0xc000402540 m=nil [GC worker (idle)]:
runtime.gopark(0x0?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000405f38 sp=0xc000405f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000405fc8 sp=0xc000405f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000405fe0 sp=0xc000405fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000405fe8 sp=0xc000405fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 10 gp=0xc000402700 m=nil [GC worker (idle)]:
runtime.gopark(0x57dd78c1be4?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000407f38 sp=0xc000407f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000407fc8 sp=0xc000407f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000407fe0 sp=0xc000407fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000407fe8 sp=0xc000407fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 11 gp=0xc0004028c0 m=nil [GC worker (idle)]:
runtime.gopark(0x57ce6b85278?, 0x0?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000411f38 sp=0xc000411f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000411fc8 sp=0xc000411f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000411fe0 sp=0xc000411fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000411fe8 sp=0xc000411fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 34 gp=0xc000484000 m=nil [GC worker (idle)]:
runtime.gopark(0x7ff688bdd9c0?, 0x1?, 0x40?, 0x1b?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00040df38 sp=0xc00040df18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc00040dfc8 sp=0xc00040df38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc00040dfe0 sp=0xc00040dfc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00040dfe8 sp=0xc00040dfe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 19 gp=0xc0001f81c0 m=nil [GC worker (idle)]:
runtime.gopark(0x57dd78c1be4?, 0x3?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00008df38 sp=0xc00008df18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc00008dfc8 sp=0xc00008df38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc00008dfe0 sp=0xc00008dfc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00008dfe8 sp=0xc00008dfe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 12 gp=0xc000402a80 m=nil [GC worker (idle)]:
runtime.gopark(0x57dd78c1be4?, 0x1?, 0xcc?, 0x2?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000413f38 sp=0xc000413f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000413fc8 sp=0xc000413f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000413fe0 sp=0xc000413fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000413fe8 sp=0xc000413fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 13 gp=0xc000402c40 m=nil [GC worker (idle)]:
runtime.gopark(0x57ce6b85278?, 0x1?, 0x88?, 0x2?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000419f38 sp=0xc000419f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000419fc8 sp=0xc000419f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000419fe0 sp=0xc000419fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000419fe8 sp=0xc000419fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 35 gp=0xc0004841c0 m=nil [GC worker (idle)]:
runtime.gopark(0x57ce6b85278?, 0x1?, 0xa8?, 0x1c?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00040ff38 sp=0xc00040ff18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc00040ffc8 sp=0xc00040ff38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc00040ffe0 sp=0xc00040ffc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00040ffe8 sp=0xc00040ffe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 20 gp=0xc0001f8380 m=nil [GC worker (idle)]:
runtime.gopark(0x57ce6b85278?, 0x3?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000415f38 sp=0xc000415f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000415fc8 sp=0xc000415f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000415fe0 sp=0xc000415fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000415fe8 sp=0xc000415fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 14 gp=0xc000402e00 m=nil [GC worker (idle)]:
runtime.gopark(0x57dd78c1be4?, 0x1?, 0x3c?, 0x31?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00041bf38 sp=0xc00041bf18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc00041bfc8 sp=0xc00041bf38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc00041bfe0 sp=0xc00041bfc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00041bfe8 sp=0xc00041bfe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 36 gp=0xc000484380 m=nil [GC worker (idle)]:
runtime.gopark(0x57dd8158f00?, 0x1?, 0x0?, 0x1a?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc00048bf38 sp=0xc00048bf18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc00048bfc8 sp=0xc00048bf38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc00048bfe0 sp=0xc00048bfc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc00048bfe8 sp=0xc00048bfe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 21 gp=0xc0001f8540 m=nil [GC worker (idle)]:
runtime.gopark(0x57dd77cbf78?, 0x3?, 0x0?, 0x0?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000417f38 sp=0xc000417f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000417fc8 sp=0xc000417f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000417fe0 sp=0xc000417fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000417fe8 sp=0xc000417fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 15 gp=0xc000402fc0 m=nil [GC worker (idle)]:
runtime.gopark(0x7ff688bdd9c0?, 0x1?, 0x28?, 0x2e?, 0x0?)
        runtime/proc.go:435 +0xce fp=0xc000487f38 sp=0xc000487f18 pc=0x7ff68650598e
runtime.gcBgMarkWorker(0xc0003211f0)
        runtime/mgc.go:1423 +0xe9 fp=0xc000487fc8 sp=0xc000487f38 pc=0x7ff6864b47a9
runtime.gcBgMarkStartWorkers.gowrap1()
        runtime/mgc.go:1339 +0x25 fp=0xc000487fe0 sp=0xc000487fc8 pc=0x7ff6864b4685
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc000487fe8 sp=0xc000487fe0 pc=0x7ff68650db21
created by runtime.gcBgMarkStartWorkers in goroutine 1
        runtime/mgc.go:1339 +0x105

goroutine 37 gp=0xc0001f88c0 m=nil [chan receive]:
runtime.gopark(0x30?, 0xffffffffffffffff?, 0x3?, 0x0?, 0xc001fa7798?)
        runtime/proc.go:435 +0xce fp=0xc001fa7750 sp=0xc001fa7730 pc=0x7ff68650598e
runtime.chanrecv(0xc00003a230, 0x0, 0x1)
        runtime/chan.go:664 +0x445 fp=0xc001fa77c8 sp=0xc001fa7750 pc=0x7ff6864a2d45
runtime.chanrecv1(0x7ff687d92ab6?, 0x29?)
        runtime/chan.go:506 +0x12 fp=0xc001fa77f0 sp=0xc001fa77c8 pc=0x7ff6864a28d2
github.com/ollama/ollama/runner/ollamarunner.(*Server).forwardBatch(_, {0x1, {0x7ff687fb3400, 0xc0096f8340}, {0x7ff687fc3298, 0xc0097c64f8}, {0xc00096b308, 0x200, 0x25f}, {{0x7ff687fc3298, ...}, ...}, ...})
        github.com/ollama/ollama/runner/ollamarunner/runner.go:476 +0xfa fp=0xc001fa7b58 sp=0xc001fa77f0 pc=0x7ff686b9f3fa
github.com/ollama/ollama/runner/ollamarunner.(*Server).run(0xc0004c6000, {0x7ff687fa59b0, 0xc0004c4000})
        github.com/ollama/ollama/runner/ollamarunner/runner.go:453 +0x18c fp=0xc001fa7fb8 sp=0xc001fa7b58 pc=0x7ff686b9f0ac
github.com/ollama/ollama/runner/ollamarunner.Execute.gowrap1()
        github.com/ollama/ollama/runner/ollamarunner/runner.go:1430 +0x28 fp=0xc001fa7fe0 sp=0xc001fa7fb8 pc=0x7ff686ba88e8
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc001fa7fe8 sp=0xc001fa7fe0 pc=0x7ff68650db21
created by github.com/ollama/ollama/runner/ollamarunner.Execute in goroutine 1
        github.com/ollama/ollama/runner/ollamarunner/runner.go:1430 +0x4c9

goroutine 38 gp=0xc0001f8a80 m=nil [select]:
runtime.gopark(0xc000047a08?, 0x2?, 0xc0?, 0x77?, 0xc00004786c?)
        runtime/proc.go:435 +0xce fp=0xc001687698 sp=0xc001687678 pc=0x7ff68650598e
runtime.selectgo(0xc001687a08, 0xc000047868, 0xe3c?, 0x0, 0x1?, 0x1)
        runtime/select.go:351 +0x837 fp=0xc0016877d0 sp=0xc001687698 pc=0x7ff6864e6437
github.com/ollama/ollama/runner/ollamarunner.(*Server).completion(0xc0004c6000, {0x7ff687fa3020, 0xc0002d01c0}, 0xc0003de3c0)
        github.com/ollama/ollama/runner/ollamarunner/runner.go:956 +0xc4e fp=0xc001687ac0 sp=0xc0016877d0 pc=0x7ff686ba388e
github.com/ollama/ollama/runner/ollamarunner.(*Server).completion-fm({0x7ff687fa3020?, 0xc0002d01c0?}, 0xc000047b40?)
        <autogenerated>:1 +0x36 fp=0xc001687af0 sp=0xc001687ac0 pc=0x7ff686ba8dd6
net/http.HandlerFunc.ServeHTTP(0xc0004ca000?, {0x7ff687fa3020?, 0xc0002d01c0?}, 0xc000047b60?)
        net/http/server.go:2294 +0x29 fp=0xc001687b18 sp=0xc001687af0 pc=0x7ff686825d49
net/http.(*ServeMux).ServeHTTP(0x7ff6864ab785?, {0x7ff687fa3020, 0xc0002d01c0}, 0xc0003de3c0)
        net/http/server.go:2822 +0x1c4 fp=0xc001687b68 sp=0xc001687b18 pc=0x7ff686827c44
net/http.serverHandler.ServeHTTP({0x7ff687f9eeb0?}, {0x7ff687fa3020?, 0xc0002d01c0?}, 0x1?)
        net/http/server.go:3301 +0x8e fp=0xc001687b98 sp=0xc001687b68 pc=0x7ff6868456ce
net/http.(*conn).serve(0xc0004d2090, {0x7ff687fa5978, 0xc0004c0360})
        net/http/server.go:2102 +0x625 fp=0xc001687fb8 sp=0xc001687b98 pc=0x7ff686824245
net/http.(*Server).Serve.gowrap3()
        net/http/server.go:3454 +0x28 fp=0xc001687fe0 sp=0xc001687fb8 pc=0x7ff686829b08
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc001687fe8 sp=0xc001687fe0 pc=0x7ff68650db21
created by net/http.(*Server).Serve in goroutine 1
        net/http/server.go:3454 +0x485

goroutine 1142 gp=0xc000485880 m=nil [IO wait]:
runtime.gopark(0x0?, 0xc0004c82a0?, 0x48?, 0x83?, 0xc0004c834c?)
        runtime/proc.go:435 +0xce fp=0xc001097d58 sp=0xc001097d38 pc=0x7ff68650598e
runtime.netpollblock(0x214?, 0x864a0406?, 0xf6?)
        runtime/netpoll.go:575 +0xf7 fp=0xc001097d90 sp=0xc001097d58 pc=0x7ff6864cbdf7
internal/poll.runtime_pollWait(0x17a78d7e7d8, 0x72)
        runtime/netpoll.go:351 +0x85 fp=0xc001097db0 sp=0xc001097d90 pc=0x7ff686504b25
internal/poll.(*pollDesc).wait(0x214?, 0x72?, 0x0)
        internal/poll/fd_poll_runtime.go:84 +0x27 fp=0xc001097dd8 sp=0xc001097db0 pc=0x7ff68659c0c7
internal/poll.execIO(0xc0004c82a0, 0x7ff687dd8778)
        internal/poll/fd_windows.go:177 +0x105 fp=0xc001097e50 sp=0xc001097dd8 pc=0x7ff68659d525
internal/poll.(*FD).Read(0xc0004c8288, {0xc000188041, 0x1, 0x1})
        internal/poll/fd_windows.go:438 +0x29b fp=0xc001097ef0 sp=0xc001097e50 pc=0x7ff68659e1fb
net.(*netFD).Read(0xc0004c8288, {0xc000188041?, 0xc00068e758?, 0xc001097f70?})
        net/fd_posix.go:55 +0x25 fp=0xc001097f38 sp=0xc001097ef0 pc=0x7ff686611945
net.(*conn).Read(0xc00020c090, {0xc000188041?, 0x7ff687facb50?, 0xc00029e320?})
        net/net.go:194 +0x45 fp=0xc001097f80 sp=0xc001097f38 pc=0x7ff686621065
net/http.(*connReader).backgroundRead(0xc000188030)
        net/http/server.go:690 +0x37 fp=0xc001097fc8 sp=0xc001097f80 pc=0x7ff68681e117
net/http.(*connReader).startBackgroundRead.gowrap2()
        net/http/server.go:686 +0x25 fp=0xc001097fe0 sp=0xc001097fc8 pc=0x7ff68681e045
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc001097fe8 sp=0xc001097fe0 pc=0x7ff68650db21
created by net/http.(*connReader).startBackgroundRead in goroutine 38
        net/http/server.go:686 +0xb6

goroutine 1172 gp=0xc00058ea80 m=nil [chan receive]:
runtime.gopark(0x30?, 0x247ff01?, 0x60?, 0x4?, 0xc0001adb20?)
        runtime/proc.go:435 +0xce fp=0xc0001adad8 sp=0xc0001adab8 pc=0x7ff68650598e
runtime.chanrecv(0xc0003201c0, 0x0, 0x1)
        runtime/chan.go:664 +0x445 fp=0xc0001adb50 sp=0xc0001adad8 pc=0x7ff6864a2d45
runtime.chanrecv1(0x7ff687d96d37?, 0x2c?)
        runtime/chan.go:506 +0x12 fp=0xc0001adb78 sp=0xc0001adb50 pc=0x7ff6864a28d2
github.com/ollama/ollama/runner/ollamarunner.(*Server).computeBatch(0xc0004c6000, {0x1, {0x7ff687fb3400, 0xc0096f8340}, {0x7ff687fc3298, 0xc0097c64f8}, {0xc00096b308, 0x200, 0x25f}, {{0x7ff687fc3298, ...}, ...}, ...})
        github.com/ollama/ollama/runner/ollamarunner/runner.go:645 +0x185 fp=0xc0001adef0 sp=0xc0001adb78 pc=0x7ff686ba0f05
github.com/ollama/ollama/runner/ollamarunner.(*Server).run.gowrap1()
        github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x58 fp=0xc0001adfe0 sp=0xc0001adef0 pc=0x7ff686b9f2d8
runtime.goexit({})
        runtime/asm_amd64.s:1700 +0x1 fp=0xc0001adfe8 sp=0xc0001adfe0 pc=0x7ff68650db21
created by github.com/ollama/ollama/runner/ollamarunner.(*Server).run in goroutine 37
        github.com/ollama/ollama/runner/ollamarunner/runner.go:459 +0x2cd
rax     0x0
rbx     0x4b92aff488
rcx     0x700000006
rdx     0x0
rdi     0xe06d7363
rsi     0x1
rbp     0x4
rsp     0x4b92aff360
r8      0x1
r9      0xe06d7363
r10     0x0
r11     0x17a732c2100
r12     0x0
r13     0x17ac69a07c0
r14     0x1000000
r15     0x0
rip     0x7ffe3ffd79da
rflags  0x206
cs      0x33
fs      0x53
gs      0x2b
time=2026-04-30T11:45:34.878+08:00 level=ERROR source=server.go:1624 msg="post predict" error="Post \"http://127.0.0.1:3427/completion\": read tcp 127.0.0.1:3434->127.0.0.1:3427: wsarecv: An existing connection was forcibly closed by the remote host."
[GIN] 2026/04/30 - 11:45:34 | 500 |   15.7829065s |       127.0.0.1 | POST     "/v1/chat/completions"

OS

Windows

GPU

Intel

CPU

Intel

Ollama version

0.22

extent analysis

TL;DR

Disabling the Vulkan backend by setting OLLAMA_GPU_OVERHEAD=0 and OLLAMA_NUM_GPU=0 may resolve the issue.

Guidance

The crash occurs inside ggml_backend_sched_graph_compute_async (Vulkan backend), suggesting a potential issue with the Vulkan backend or the GPU.
Disabling the Vulkan backend (forcing CPU) resolves the issue, indicating that the problem is likely related to the GPU or the Vulkan backend.
The error message "wsarecv: An existing connection was forcibly closed by the remote host" suggests a potential issue with the network connection or the server.
The stack trace shows that the crash occurs in the computeBatch function of the ollamarunner package, which is responsible for computing the batch of requests.

Example

No code snippet is provided as the issue is related to a specific hardware and software configuration.

Notes

The issue is specific to the Intel ARC A750 GPU and the Vulkan backend, and may be related to a driver or hardware issue. Further investigation is needed to determine the root cause of the problem.

Recommendation

Apply workaround: Disable the Vulkan backend by setting OLLAMA_GPU_OVERHEAD=0 and OLLAMA_NUM_GPU=0 to force the use of the CPU backend. This may resolve the issue, but may also impact performance.

Vote matrix · Quick signals

Works

Did the solution work? Tap to confirm.

Easy Fix

Was it a quick fix?

Time Saver

Did it save you time?

Blocking

Was it severely blocking?

Common Issue

Are others likely hitting this too?

Flaky / Intermittent

Is it intermittent?

Verified / Reproducible

Can you reproduce it reliably?

#api #model loading #serialization error #model compatibility #GPU setup

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Data

Security

Network

Code

UI/UX

Text

System

Multimedia

Protocol

API

Engineering

ollama - 💡(How to fix) Fix Vulkan backend crashes with C++ exception on Intel ARC A750 when using gemma4 26b MoE + image input [1 participants]

Recommended Tools

GitHub issue graph ai analysis

Error Message

Root Cause

Fix Action

Fix / Workaround

Code Example

What is the issue?

Relevant log output

OS

GPU

CPU

Ollama version

extent analysis

TL;DR

Guidance

Example

Notes

Recommendation

Still need to ship something?

TRENDING

ollama - 💡(How to fix) Fix Vulkan backend crashes with C++ exception on Intel ARC A750 when using gemma4 26b MoE + image input [1 participants]

Recommended Tools

GitHub issue graph ai analysis

Error Message

Root Cause

Fix Action

Fix / Workaround

Code Example

What is the issue?

Relevant log output

OS

GPU

CPU

Ollama version

extent analysis

TL;DR

Guidance

Example

Notes

Recommendation

Still need to ship something?

RELATED_DISCOVERY

TRENDING