-
Update README.md
-
Upload 3 files
# Models
This directory stores model weights in two clearly separated trees:
```
models/
├── pytorch/ # original PyTorch weights (HF diffusers format, untouched)
│ ├── stable-diffusion-v1-5/ # unet/ vae/ text_encoder/
│ ├── stable-diffusion-xl-base-1.0/ # unet/ vae/ text_encoder/ text_encoder_2/
│ ├── brushnet-sd1.5-seg/ # segmentation-mask brushnet checkpoint
│ └── brushnet-sdxl-seg/
└── gguf/ # converted GGUF weights (every component x precision)
├── sd1.5/ text_encoder|unet|vae|brushnet/{f32,f16,q8_0}/
└── sdxl/ text_encoder|text_encoder_2|unet|vae|brushnet/{f32,f16,q8_0}/
```
## Naming convention
- GGUF files: `<component>.<precision>.gguf`, e.g. `unet.q8_0.gguf`.
- Precisions: `f32` (PyTorch-parity reference), `f16` (16-bit float),
`q8_0` (8-bit block quantization).
## Workflow
1. Download the PyTorch checkpoints into `pytorch/`:
`bash scripts/download_models.sh`
(or place … -
initial commit
