mudler
stablediffusion fails to install on DGX Spark
LocalAI version: sha-ef44ace-nvidia-l4t-arm64-cuda-13
Environment, CPU architecture, OS, and Version: docker
-> % lscpu
Architecture: aarch64
CPU op-mode(s): 64-bit
Byte Order: Little Endian
CPU(s): 20
On-line CPU(s) list: 0-19
Vendor ID: ARM
Model name: Cortex-X925
Model: 1
Thread(s) per core: 1
Core(s) per socket: 10
Socket(s): 1
Stepping: r0p1
CPU(s) scaling MHz: 90%
CPU max MHz: 4004.0000
CPU min MHz: 1378.0000
BogoMIPS: 2000.00
Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fc
ma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca
pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16
i8mm bf16 dgh bti ecv afp wfxt
Model name: Cortex-A725
Model: 1
Thread(s) per core: 1
Core(s) per socket: 10
Socket(s): 1
Stepping: r0p1
CPU(s) scaling MHz: 86%
CPU max MHz: 2860.0000
CPU min MHz: 338.0000
BogoMIPS: 2000.00
Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fc
ma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca
pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16
i8mm bf16 dgh bti ecv afp wfxt
-> % lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description: Ubuntu 24.04.3 LTS
Release: 24.04
Codename: noble
-> % docker exec -it local-ai bash
root@5837be385699:/# nvidia-smi
Wed Dec 10 13:14:41 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.95.05 Driver Version: 580.95.05 CUDA Version: 13.0 |
+-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA GB10 On | 0000000F:01:00.0 Off | N/A |
| N/A 39C P8 4W / N/A | Not Supported | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| No running processes found |
+-----------------------------------------------------------------------------------------+
Describe the bug stablediffusion backend fails to install with;
Error installing backend "cuda13-nvidia-l4t-arm64-stablediffusion-ggml": not a valid backend: run file not found "/backends/cuda13-nvidia-l4t-arm64-stablediffusion-ggml/run.sh"
To Reproduce docker run -ti --name local-ai -p 32000:8080 --gpus all localai/localai:sha-ef44ace-nvidia-l4t-arm64-cuda-13
navigate to the local site > backends > try to install cuda13-nvidia-l4t-arm64-stablediffusion-ggml
Expected behavior the backend should install
Logs 1:08PM ERR Run file not found runFile=/backends/cuda13-nvidia-l4t-arm64-stablediffusion-ggml/run.sh 1:08PM ERR error installing backend localai@cuda13-nvidia-l4t-arm64-stablediffusion-ggml error="not a valid backend: run file not found "/backends/cuda13-nvidia-l4t-arm64-stablediffusion-ggml/run.sh""
Additional context
cuda 13 support on DGX spark is in the works (but it's functional): you need to pull the corresponding development backend for now until there is a tagged release.
cuda 13 support on DGX spark is in the works (but it's functional): you need to pull the corresponding
developmentbackend for now until there is a tagged release.
thank you! i will give that a shot
that worked! however the generated images via dreamshaper are not at all what the prompt is lol. i suppose this is why its still in development. anyway i will periodically check back on progress. thanks again for replying
ive just begun poking around with stable diffusion so this could be a me issue if this should have worked?
"A cyberpunk cityscape with neon lights reflecting on rainy streets, people wearing futuristic clothing, ultra-detailed."
cuda13-nvidia-l4t-arm64-diffusers-development
cuda13-nvidia-l4t-arm64-stablediffusion-ggml-development
things are much better now - can you give it a shot again with latest release? I've tried so far with qwen-image, but looking good so far.
The image that should be used for the DGX spark is: localai/localai:latest-nvidia-l4t-arm64-cuda-13
im not getting any errors with qwen3-image but it appears to not be running at all and nothing of note in the logs.
i did have to set f16: false in the model config. it was complaining about no diffusion pipeline; after prompting however it just sits there and i dont see any load in nvtop or htop either.
maybe im missing something not sure.
however with dreamshaper i am seeing a significant improvement. im not sure why but the first attempt produced undesirable results. not as bad as my previous screenshot but not much better either.
nonetheless, the great news is my 2nd and 3rd attempt with dreamshaper produced something pretty awesome;
A cyberpunk cityscape with neon lights reflecting on rainy streets, people wearing futuristic clothing, ultra-detailed.
cuda13-nvidia-l4t-arm64-stablediffusion-ggml-development
LocalAI Version v3.9.0 (aadec0b8cb2a7c608981823e7b8a003551662205)
im not getting any errors with
qwen3-imagebut it appears to not be running at all and nothing of note in the logs.
Note that this could be just the model being downloaded in the background. In the case of qwen-image the first request will cause the model to download if not already present, and qwen image is quite big (IIRC around 40gb).
However great to hear that now dreamshaper works!
Note that this could be just the model being downloaded in the background.
good call. tried again this morning
qwen-image