These instructions are based on DreamBooth usage with the https://github.com/ShivamShrirao/diffusers repo.
Below are 2 files. "convertToCkpt.py" and "toCkpt.sh". Create those files inside the examples/dreambooth folder with the code provided.
| { | |
| "LoRA_type": "Standard", | |
| "adaptive_noise_scale": 0, | |
| "additional_parameters": "", | |
| "block_alphas": "", | |
| "block_dims": "", | |
| "block_lr_zero_threshold": "", | |
| "bucket_no_upscale": true, | |
| "bucket_reso_steps": 64, | |
| "cache_latents": true, |
These instructions are based on DreamBooth usage with the https://github.com/ShivamShrirao/diffusers repo.
Below are 2 files. "convertToCkpt.py" and "toCkpt.sh". Create those files inside the examples/dreambooth folder with the code provided.
| # Script for converting a HF Diffusers saved pipeline to a Stable Diffusion checkpoint. | |
| # *Only* converts the UNet, VAE, and Text Encoder. | |
| # Does not convert optimizer state or any other thing. | |
| # Written by jachiam | |
| import argparse | |
| import os.path as osp | |
| import torch |