---
library_name: diffusers
pipeline_tag: text-to-image
---
**Warning: THIS model is NOT suitable for use by minors. The model will generate X-rated/NFSW content.**

# E621 Rising Stable Diffusion 2.1 Model [epoch 19]
* Guaranteed **NSFW** or your money back
* Fine-tuned from [Stable Diffusion v2-1-base](https://huggingface.co/stabilityai/stable-diffusion-2-1-base)
* 19 epochs of 450,000 images each, collected from [E621](https://e621.net/) and curated based on scores, favorite counts, and tag filtering.
* Trained with [5,356 tags](./meta/tag-counts.json)
* `512x512px`
* Compatible with 🤗 `diffusers`
* Compatible with `stable-diffusion-webui`
* Likely compatible with anything that accepts [`.ckpt` and `.yaml` files](https://huggingface.co/hearmeneigh/sd21-e621-rising-v1-ckpt/tree/main)

## Getting Started
* [Stable Diffusion WebUI How-To](./guides/WEBUI.md)
* [Python How-To](./guides/PYTHON.md)

## Example Prompt
```
anthro solo female standing rating:questionable

species:equine biped
two_tone_fur grey_body grey_fur white_fur white_snout white_markings gloves_marking white_tail
blue_eyes facial_markings white_hair white_mane evil_grin 
athletic_female

meta:shaded
meta:digital_media_artwork
meta:detailed
meta:digital_painting_artwork

seductive looking_at_viewer
tomboy
tomb raider outfit
```

## Changes From E621
See a [complete list of tags here](./meta/tag-counts.json).

* Symbols have been prefixed with `symbol:`, e.g. `symbol:<3`
* All categories except `general` have been prefixed with the category name, e.g. `copyright:somename`. The categories are:
  * `artist`
  * `copyright`
  * `character`
  * `species`
  * `invalid`
  * `meta`
  * `lore`
* Tag names are all lowercase and only contain `a-z`, `0-9`, `/`, and `_` letters
* `:` is used to separate the category name from the tag

## Training Procedure
* 204-272 images per batch (epoch variant)
* `512x512px` image size
* Adam optimizer
  * Beta1 = `0.9`
  * Beta2 = `0.999`
  * Weight decay = `1e-2`
  * Epsilon = `1e-08`
* Constant learning rate `4e-6`
* `bf16` mixed precision
* 12 epochs of samples stretched to `512x512px` (ignore aspect ratio)
* 4 epochs of samples resized to `512xH` or `Wx512px` with center crop (maintain aspect ratio)
* 3 epochs of samples resized to `< 512x512px` (maintain aspect ratio)
* Tags for each sample are shuffled for each epoch, starting from epoch 16