Cuda error

[razvanab]

Cuda error

.\sd.exe -m ".\SDXL.safetensors" --lora-model-dir ".\loras\SDXL" --clip-skip 2 --seed -1 --steps 25 --cfg-scale 7.5 -H 768 -W 768 --schedule exponential --sampling-method euler_a -p "a woman in red in the park" --negative-prompt " messy drawing, amateur drawing, ugly face, bad face, bad teeth, (interlocked fingers, extra fingers, badly drawn hands and fingers, anatomically incorrect hands, bad anatomy), (blurry, depth of field) ,sketch, painting, simple background, white background" --vae ".\Vae\fp16_fix_sdxl_vae.safetensors"

GPU: RTX 3050 8GB Vram
Cuda toolkit 12.4

PyTorch version: 2.6.0+cu124
CUDA toolkit version:
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2024 NVIDIA Corporation
Built on Tue_Feb_27_16:28:36_Pacific_Standard_Time_2024
Cuda compilation tools, release 12.4, V12.4.99
Build cuda_12.4.r12.4/compiler.33961263_0

ggml_cuda_compute_forward: GET_ROWS failed
CUDA error: no kernel image is available for execution on the device
  current device: 0, in function ggml_cuda_compute_forward at D:\a\stable-diffusion.cpp\stable-diffusion.cpp\ggml\src\ggml-cuda\ggml-cuda.cu:2174
  err
D:\a\stable-diffusion.cpp\stable-diffusion.cpp\ggml\src\ggml-cuda\ggml-cuda.cu:70: CUDA error

[fszontagh]

If I'm right, this is a precompiled version of sd.exe.

The error you encountered:

"CUDA error: no kernel image is available for execution on the device"

This means your GPU's architecture is not supported. You need to compile it yourself or You can try to run the Linux version in WSL.

The "PyTorch version: 2.6.0+cu124" information is not relevant here, as there are no Python dependencies.

[razvanab]

I see.
Thanks