---
pipeline_tag: text-to-image
widget:
- text: Three cow grazing in a bay window
  output:
    url: cow.png
- text: >-
    Super Closeup Portrait, action shot, Profoundly dark whitish meadow, glass
    flowers, Stains, space grunge style, Jeanne d'Arc wearing White Olive green
    used styled Cotton frock, Wielding thin silver sword, Sci-fi vibe, dirty,
    noisy, Vintage monk style, very detailed, hd
  output:
    url: girl.png
- text: >-
    spacious,circular underground room,{dirtied and bloodied white
    tiles},amalgamation,flesh,plastic,dark fabric,core,pulsating
    heart,limbs,human-like arms,twisted angelic wings,arms,covered in
    skin,feathers,scales,undulate slowly,unseen current,convulsing,head
    area,chaotic,mass of eyes,mouths,no human features,smaller
    forms,cherubs,demons,golden wires,surround,holy light,tv static
    effect,golden glow,shadows,terrifying essence,overwhelming
    presence,nightmarish,landscape,sparse,cavernous,eerie,dynamic,motion,striking,awe-inspiring,nightmarish,nightmarish,nightmare,horrifying,bio-mechanical,body
    horror,amalgamation
  output:
    url: aigle.png
license: gpl-3.0
datasets:
- CortexLM/midjourney-v6
library_name: diffusers
language:
- en
tags:
- bittensor
- decentralization
- vision
- subnet 19
- pixel
---
<img src="https://cdn-uploads.huggingface.co/production/uploads/655a0bdf3ff5ba1b1b1c01b7/y1dKBZh8UhII6wtbs5boj.png" alt="drawing" width="372"/>

<Gallery />

(From <a href="https://huggingface.co/PlixAI/BitDiffusionV0.1">PlixAI BitDiffusion model</a>)

### BitDiffusionXL v0.1

This is the initial version of the image model trained on the Bittensor network within Pixel subnet. It's not expected for this model to perform as well as MidJourney V6 at the moment. However, it does generate better images than base SDXL model.

**Trained on the dataset of Subnet 19 Vision.**

## Pixel subnet Checkpoint

Model ID : gtsru/sn17-dek-012

Revision : 5852d39e8413a377a3477b8278ade9af311f83a4

UID : 42

Perplexity : 1.1325

## Settings for BitDiffusionXL v0.1

Use these settings for the best results with BitDiffusionV0.1:

CFG Scale: Use a CFG scale of 8

Steps: 40 to 60 steps

Sampler: DPM++ 2M SDE

Scheduler: Karras

Resolution: 1024x1024

**For best results, set a negative_prompt**

## Use it with 🧨 diffusers
```python
import torch
from diffusers import (
    StableDiffusionXLPipeline, 
    KDPM2AncestralDiscreteScheduler,
    AutoencoderKL
)

# Load VAE component
vae = AutoencoderKL.from_pretrained(
    "madebyollin/sdxl-vae-fp16-fix", 
    torch_dtype=torch.float16
)

# Configure the pipeline
pipe = StableDiffusionXLPipeline.from_pretrained(
    "CortexLM/BitDiffusionXL-v0.1", 
    vae=vae,
    torch_dtype=torch.float16
)
pipe.scheduler = KDPM2AncestralDiscreteScheduler.from_config(pipe.scheduler.config)
pipe.to('cuda')

# Define prompts and generate image
prompt = "black fluffy gorgeous dangerous cat animal creature, large orange eyes, big fluffy ears, piercing gaze, full moon, dark ambiance, best quality, extremely detailed"
negative_prompt = "nsfw, bad quality, bad anatomy, worst quality, low quality, low resolutions, extra fingers, blur, blurry, ugly, wrongs proportions, watermark, image artifacts, lowres, ugly, jpeg artifacts, deformed, noisy image"

image = pipe(
    prompt, 
    negative_prompt=negative_prompt, 
    width=1024,
    height=1024,
    guidance_scale=7.5,
    num_inference_steps=50
).images[0]
```

Training Subnet : https://github.com/PlixML/pixel

Dataset : https://huggingface.co/datasets/CortexLM/midjourney-v6