|
--- |
|
license: openrail++ |
|
language: |
|
- en |
|
tags: |
|
- stable-diffusion |
|
- sygil-diffusion |
|
- text-to-image |
|
- sygil-devs |
|
- finetune |
|
- stable-diffusion-1.5 |
|
inference: true |
|
pinned: true |
|
--- |
|
|
|
|
|
# About the model |
|
----------------- |
|
This model is a Stable Diffusion v1.5 fine-tune trained on the [Imaginary Network Expanded Dataset](https://github.com/Sygil-Dev/INE-dataset). |
|
It is an advanced version of Stable Diffusion and can generate nearly all kinds of images like humans, reflections, cities, architecture, fantasy, concepts arts, anime, manga, digital arts, landscapes, or nature views. |
|
This model allows the user to have total control of the generation as they can use multiple tags and namespaces to control almost everything |
|
on the final result including image composition. |
|
|
|
**Note that the prompt engineering techniques is a bit different from other models and Stable Diffusion, |
|
while you can still use normal prompts like in other Stable Diffusion modelsin order to get the best out of this model you will need to make use of tags and namespaces. |
|
More information about namespace will later be added. |
|
<!--- |
|
[More about it here](promptingGuide.md)** \ |
|
--> |
|
\ |
|
**If you find our work useful, please consider supporting us using one of the options below:** |
|
- [OpenCollective](https://opencollective.com/sygil_dev) |
|
|
|
**Join our Discord Server for supports and announcements** |
|
[](https://discord.gg/fTtcufxyHQ) |
|
|
|
# Showcase |
|
 |
|
|
|
## Examples |
|
|
|
Using the [🤗's Diffusers library](https://github.com/huggingface/diffusers) to run Sygil Diffusion in a simple and efficient manner. |
|
|
|
```bash |
|
pip install diffusers transformers accelerate scipy safetensors |
|
``` |
|
Running the pipeline (if you don't swap the scheduler it will run with the default DDIM, in this example we are swapping it to DPMSolverMultistepScheduler): |
|
|
|
```python |
|
from diffusers import StableDiffusionPipeline, DPMSolverMultistepScheduler |
|
|
|
model_id = "Sygil/Sygil-Diffusion" |
|
|
|
# Use the DPMSolverMultistepScheduler (DPM-Solver++) scheduler here instead |
|
pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16) |
|
pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config) |
|
pipe = pipe.to("cuda") |
|
|
|
prompt = "a photo of an astronaut riding a horse on mars" |
|
image = pipe(prompt).images[0] |
|
|
|
image.save("astronaut_rides_horse.png") |
|
``` |
|
|
|
**Notes**: |
|
- Despite not being a dependency, we highly recommend you to install [xformers](https://github.com/facebookresearch/xformers) for memory efficient attention (better performance) |
|
- If you have low GPU RAM available, make sure to add a `pipe.enable_attention_slicing()` after sending it to `cuda` for less VRAM usage (to the cost of speed). |
|
|
|
|
|
## Training |
|
|
|
**Training Data** |
|
The model was trained on the following dataset: |
|
- [Imaginary Network Expanded Dataset](https://github.com/Sygil-Dev/INE-dataset) dataset. |
|
|
|
**Hardware and others** |
|
- **Hardware:** 1 x Nvidia RTX 3050 8GB GPU |
|
- **Hours Trained:** 384 approximately. |
|
- **Optimizer:** AdamW |
|
- **Gradient Accumulations**: 1 |
|
- **Batch:** 1 |
|
- **Learning rate:** warmup to 1e-7 for 10,000 steps and then kept constant |
|
- **Total Training Steps:** 800,0000 |
|
|
|
Developed by: [Sygil-Dev](https://github.com/Sygil-Dev/) |
|
|
|
# License |
|
This model is open access and available to all, with a CreativeML Open RAIL++-M License further specifying rights and usage. [Please read the full license here](https://huggingface.co/stabilityai/stable-diffusion-2/blob/main/LICENSE-MODEL) |
