whisper breaks on cuda-13

[markuman]

LocalAI version:

localai/localai:v3.9.0-aio-gpu-nvidia-cuda-13

Environment, CPU architecture, OS, and Version:

Linux gpu2 6.18.5-arch1-1 #1 SMP PREEMPT_DYNAMIC Sun, 11 Jan 2026 17:10:53 +0000 x86_64 GNU/Linux

Describe the bug

Jan 14 12:56:08 ERROR failed starting/connecting to the gRPC service error=rpc error: code = Unavailable desc = connection error: desc = "transport: Error while dialing: dial tcp 127.0.0.1:40857: connect: connection refused" 
Jan 14 12:56:10 ERROR Failed to load model modelID="ggml-large-v3-turbo.bin" error=failed to load model with internal loader: grpc service not ready backend="whisper" 

using

localai/localai:v3.9.0-aio-gpu-nvidia-cuda-12

fixes the problem

[mudler]

@markuman please share the full logs with --debug true

[markuman]

@mudler

Jan 15 15:11:28 INFO  LocalAI is started and running address=":8080" caller={caller.file="/build/core/cli/run.go"  caller.L=254 } 
Jan 15 15:11:59 DEBUG overriding empty model name in request body with value found earlier in middleware chain context localModelName="ggml-large-v3-turbo.bin" caller={caller.file="/build/core/http/middleware/request.go"  caller.L=138 } 
Jan 15 15:11:59 DEBUG input.Input input="<nil>" caller={caller.file="/build/core/http/middleware/request.go"  caller.L=412 } 
Jan 15 15:11:59 DEBUG Audio file copied dst="/tmp/whisper2913994183/lekker.mp3" caller={caller.file="/build/core/http/endpoints/openai/transcription.go"  caller.L=71 } 
Jan 15 15:11:59 INFO  BackendLoader starting modelID="ggml-large-v3-turbo.bin" backend="whisper" model="ggml-large-v3-turbo.bin" caller={caller.file="/build/pkg/model/initializers.go"  caller.L=159 } 
Jan 15 15:11:59 DEBUG Loading model in memory from file file="/models/ggml-large-v3-turbo.bin" caller={caller.file="/build/pkg/model/loader.go"  caller.L=204 } 
Jan 15 15:11:59 DEBUG Loading Model with gRPC modelID="ggml-large-v3-turbo.bin" file="/models/ggml-large-v3-turbo.bin" backend="whisper" options={whisper ggml-large-v3-turbo.bin ggml-large-v3-turbo.bin {{}} 0xc00041f208 map[] 20 2 false} caller={caller.file="/build/pkg/model/initializers.go"  caller.L=53 } 
Jan 15 15:11:59 DEBUG Loading external backend uri="/backends/cuda12-whisper/run.sh" caller={caller.file="/build/pkg/model/initializers.go"  caller.L=77 } 
Jan 15 15:11:59 DEBUG external backend is file file=&{run.sh 1223 493 {0 63899791382 0x4e11ce0} {64514 1572870 1 33261 65534 65534 0 0 1223 4096 8 {1768395330 226581130} {1764194582 0} {1768395155 337749033} [0 0 0]}} caller={caller.file="/build/pkg/model/initializers.go"  caller.L=80 } 
Jan 15 15:11:59 DEBUG Loading GRPC Process process="/backends/cuda12-whisper/run.sh" caller={caller.file="/build/pkg/model/process.go"  caller.L=105 } 
Jan 15 15:11:59 DEBUG GRPC Service will be running id="ggml-large-v3-turbo.bin" address="127.0.0.1:36141" caller={caller.file="/build/pkg/model/process.go"  caller.L=107 } 
Jan 15 15:11:59 DEBUG GRPC Service state dir dir="/tmp/go-processmanager3077359356" caller={caller.file="/build/pkg/model/process.go"  caller.L=131 } 
Jan 15 15:11:59 DEBUG GRPC Service Started caller={caller.file="/build/pkg/model/initializers.go"  caller.L=92 } 
Jan 15 15:11:59 DEBUG Wait for the service to start up caller={caller.file="/build/pkg/model/initializers.go"  caller.L=105 } 
Jan 15 15:11:59 DEBUG Options options=ContextSize:1024  Seed:626776370  NBatch:512  MMap:true  NGPULayers:9999999  Threads:10  FlashAttention:"auto" caller={caller.file="/build/pkg/model/initializers.go"  caller.L=106 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+++ realpath run.sh" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="++ dirname /backends/cuda12-whisper/run.sh" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ CURDIR=/backends/cuda12-whisper" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ cd /" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ echo 'CPU info:'" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stdout id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="CPU info:" caller={caller.file="/build/pkg/model/process.go"  caller.L=155 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="++ uname" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ '[' Linux '!=' Darwin ']'" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ grep -e 'model\\sname' /proc/cpuinfo" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ head -1" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stdout id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="model name\t: 13th Gen Intel(R) Core(TM) i5-13500" caller={caller.file="/build/pkg/model/process.go"  caller.L=155 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ grep -e flags /proc/cpuinfo" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ head -1" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stdout id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="flags\t\t: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid rdseed adx smap clflushopt clwb intel_pt sha_ni xsaveopt xsavec xgetbv1 xsaves split_lock_detect user_shstk avx_vnni dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp hwp_pkg_req hfi vnmi umip pku ospke waitpkg gfni vaes vpclmulqdq rdpid movdiri movdir64b fsrm md_clear serialize pconfig arch_lbr ibt flush_l1d arch_capabilities" caller={caller.file="/build/pkg/model/process.go"  caller.L=155 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ LIBRARY=/backends/cuda12-whisper/libgowhisper-fallback.so" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="++ uname" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ '[' Linux '!=' Darwin ']'" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ grep -q -e '\\savx\\s' /proc/cpuinfo" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stdout id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="CPU:    AVX    found OK" caller={caller.file="/build/pkg/model/process.go"  caller.L=155 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ echo 'CPU:    AVX    found OK'" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ '[' -e /backends/cuda12-whisper/libgowhisper-avx.so ']'" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ LIBRARY=/backends/cuda12-whisper/libgowhisper-avx.so" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ grep -q -e '\\savx2\\s' /proc/cpuinfo" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stdout id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="CPU:    AVX2   found OK" caller={caller.file="/build/pkg/model/process.go"  caller.L=155 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ echo 'CPU:    AVX2   found OK'" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ '[' -e /backends/cuda12-whisper/libgowhisper-avx2.so ']'" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ LIBRARY=/backends/cuda12-whisper/libgowhisper-avx2.so" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ grep -q -e '\\savx512f\\s' /proc/cpuinfo" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ export LD_LIBRARY_PATH=/backends/cuda12-whisper/lib:" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ LD_LIBRARY_PATH=/backends/cuda12-whisper/lib:" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ export WHISPER_LIBRARY=/backends/cuda12-whisper/libgowhisper-avx2.so" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ WHISPER_LIBRARY=/backends/cuda12-whisper/libgowhisper-avx2.so" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ '[' -f /backends/cuda12-whisper/lib/ld.so ']'" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ echo 'Using lib/ld.so'" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ echo 'Using library: /backends/cuda12-whisper/libgowhisper-avx2.so'" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="+ exec /backends/cuda12-whisper/lib/ld.so /backends/cuda12-whisper/whisper --addr 127.0.0.1:36141" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stdout id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="Using lib/ld.so" caller={caller.file="/build/pkg/model/process.go"  caller.L=155 } 
Jan 15 15:11:59 DEBUG GRPC stdout id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="Using library: /backends/cuda12-whisper/libgowhisper-avx2.so" caller={caller.file="/build/pkg/model/process.go"  caller.L=155 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="panic: libcudart.so.12: cannot open shared object file: No such file or directory" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="goroutine 1 [running]:" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="main.main()" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:11:59 DEBUG GRPC stderr id="ggml-large-v3-turbo.bin-127.0.0.1:36141" line="\t/LocalAI/backend/go/whisper/main.go:30 +0x274" caller={caller.file="/build/pkg/model/process.go"  caller.L=146 } 
Jan 15 15:12:23 INFO  HTTP request method="GET" path="/readyz" status=200 caller={caller.file="/build/core/http/app.go"  caller.L=111 } 
Jan 15 15:12:37 ERROR failed starting/connecting to the gRPC service error=rpc error: code = Unavailable desc = connection error: desc = "transport: Error while dialing: dial tcp 127.0.0.1:36141: connect: connection refused" caller={caller.file="/build/pkg/model/initializers.go"  caller.L=118 } 
Jan 15 15:12:39 DEBUG GRPC Service NOT ready caller={caller.file="/build/pkg/model/initializers.go"  caller.L=124 } 
Jan 15 15:12:39 ERROR Failed to load model modelID="ggml-large-v3-turbo.bin" error=failed to load model with internal loader: grpc service not ready backend="whisper" caller={caller.file="/build/pkg/model/initializers.go"  caller.L=179 } 
Jan 15 15:12:39 INFO  HTTP request method="POST" path="/v1/audio/transcriptions" status=200 caller={caller.file="/build/core/http/app.go"  caller.L=111 } 

running with

docker run -ti --rm --name localai -e LOCALAI_DISABLE_WEBUI=true -e LOCALAI_UPLOAD_LIMIT=512 -e LOCALAI_FORCE_META_BACKEND_CAPABILITY=nvidia -v /opt/localai/models/:/models:cached -v /opt/localai/backends/:/backends:cached -p 127.0.0.1:8080:8080 --device=nvidia.com/gpu=all docker.io/localai/localai:v3.9.0-aio-gpu-nvidia-cuda-13 --debug true
===> LocalAI All-in-One (AIO) container starting...
NVIDIA GPU detected
Thu Jan 15 15:10:23 2026       
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 590.48.01              Driver Version: 590.48.01      CUDA Version: 13.1     |
+-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA RTX 4000 SFF Ada ...    Off |   00000000:01:00.0 Off |                  Off |
| 30%   45C    P8             12W /   70W |       2MiB /  20475MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|  No running processes found                                                             |
+-----------------------------------------------------------------------------------------+
NVIDIA GPU detected. Attempting to find memory size...
Total GPU Memory: 20475 MiB
...