v1.0.1
sd-cli-cuda v1.0.1
CUDA-accelerated Stable Diffusion backend for wavespeed-desktop.
What's New
- CUDA memory check - Requires 8GB free GPU memory before running
- Exits immediately with clean "CUDA Out of Memory!" message when memory is low
- Prevents silent failures that result in empty/blank images
- Displays properly in wavespeed-desktop as "Generation failed: CUDA Out of Memory!"
- CUDA runtime dependencies - Package declares proper dependencies
- Auto-installs
libcudart12,libcublas12,libcublaslt12when using apt
- Auto-installs
Installation
# Recommended: auto-installs CUDA dependencies
sudo apt install ./sd-cli-cuda_1.0.1_amd64.deb
# Or manually (requires CUDA libs already installed)
sudo dpkg -i sd-cli-cuda_1.0.1_amd64.debRequirements
- Debian-based distro (Ubuntu 24.04 LTS recommended)
- NVIDIA GPU with CUDA support (8GB+ VRAM free required)
- NVIDIA drivers installed (nvidia-driver-580 or newer for CUDA 12+)
Troubleshooting
If you see "CUDA Out of Memory!", free up GPU memory by closing GPU-intensive apps:
sudo pkill -f 'main.py' # ComfyUI