Vixy
e4294b57e6
- SDXL image generation using RealVisXL_V4.0 - IP-Adapter FaceID integration for consistent face generation - Simplified API (removed client_id requirement) - New params: face_image, face_strength - 'vixy' shortcut for face-locked generation - Queue-based async job processing - FastAPI with proper error handling Co-authored-by: Alex <alex@k4zka.online>
54 lines
1.3 KiB
Bash
Executable File
54 lines
1.3 KiB
Bash
Executable File
#!/bin/bash
|
|
# Download SDXL models for DreamTail using Docker
|
|
set -e
|
|
|
|
echo "📥 Downloading SDXL models..."
|
|
echo "This will download ~13GB of model weights"
|
|
echo ""
|
|
|
|
# Model cache directory
|
|
MODELS_DIR="${DREAMTAIL_MODELS:-/mnt/nvme/models}"
|
|
|
|
# Create directory if it doesn't exist
|
|
mkdir -p "$MODELS_DIR"
|
|
|
|
echo "Models will be cached in: $MODELS_DIR"
|
|
echo ""
|
|
echo "Using Docker container to download models..."
|
|
echo ""
|
|
|
|
# Use L4T PyTorch container to download models
|
|
docker run --rm -it \
|
|
-v "${MODELS_DIR}:/models" \
|
|
dustynv/l4t-pytorch:r36.2.0-pth2.1-py3 \
|
|
bash -c "
|
|
pip3 install -q diffusers transformers accelerate safetensors &&
|
|
python3 << 'PYEOF'
|
|
from diffusers import StableDiffusionXLPipeline
|
|
|
|
model_id = 'stabilityai/stable-diffusion-xl-base-1.0'
|
|
cache_dir = '/models'
|
|
|
|
print(f'Downloading {model_id}...')
|
|
print(f'Cache directory: {cache_dir}')
|
|
print('')
|
|
|
|
try:
|
|
pipeline = StableDiffusionXLPipeline.from_pretrained(
|
|
model_id,
|
|
use_safetensors=True,
|
|
cache_dir=cache_dir
|
|
)
|
|
print('✅ SDXL model downloaded successfully!')
|
|
except Exception as e:
|
|
print(f'❌ Error downloading model: {e}')
|
|
exit(1)
|
|
PYEOF
|
|
"
|
|
|
|
echo ""
|
|
echo "✅ Model download complete!"
|
|
echo ""
|
|
echo "Models are cached in: $MODELS_DIR"
|
|
echo "You can now build and run DreamTail"
|