license: creativeml-openrail-m
datasets:
- ChristophSchuhmann/improved_aesthetics_6plus
- drhead/laion_hd_21M_deduped
pipeline_tag: text-to-image
Currently released models: ZeroDiffusion-Base v0.9 (zd_base_v0-9 and zd_base_v0-9_ema)
Currently training model: ZeroDiffusion-Inpaint, a finetuned inpainting model with zero terminal SNR trained on synthetic masks.
This is a work in progress model trained off of SD 1.5 with zero terminal SNR.
ZeroDiffusion v0.9 is intended as a final prototype made from a complete training run. ZeroDiffusion v1.0 will involve another full restart from Stable Diffusion v1.5.
The intention of this model is to provide a training base but I politely ask that you do not make any major training runs you intend to release on the prototype epochs. Please wait for the full release for that to maximize compatibility. Experiment all you'd like though!
For this model to work well, you will probably need CFG rescale and for the DDIM sampler to use a trailing timestep selection. Both are implemented in this plugin: https://github.com/Seshelle/CFG_Rescale_webui
You must also download the corresponding YAML file and put it in the folder with the model (assuming you are using A1111's webui or similar). It won't work without it. It will tell webui to use the model in v-prediction mode.