SDXL LoRA DreamBooth - zamasW/drug_resized_LoRA

Model description

These are zamasW/drug_resized_LoRA LoRA adaption weights for stabilityai/stable-diffusion-xl-base-1.0.

The weights were trained using DreamBooth.

LoRA for the text encoder was enabled: False.

Special VAE used for training: madebyollin/sdxl-vae-fp16-fix.

Trigger words

You should use a face after drug abuse to trigger the image generation.

Download model

Weights for this model are available in Safetensors format.

Download them in the Files & versions tab.

Intended uses & limitations

How to use

import torch
from diffusers import DiffusionPipeline, AutoencoderKL
repo_id = 'zamasW/drug_resized_LoRA'
vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
pipe = DiffusionPipeline.from_pretrained(
    "stabilityai/stable-diffusion-xl-base-1.0",
    vae=vae,
    torch_dtype=torch.float16,
    variant="fp16",
    use_safetensors=True
)
pipe.load_lora_weights(repo_id)
_ = pipe.to("cuda")

prompt = "a face after drug abuse" 

image = pipe(prompt=prompt, num_inference_steps=25).images[0]

Limitations and bias

coming soon

Training details

Dataset : over 50 images of 1080*1080 of drug abused faces

• Model: stabilityai/stable-diffusion-xl-base-1.0
• VAE: madebyollin/sdxl-vae-fp16-fix
• Instance Prompt: "a face after dug abuse"
• Image Resolution: 1024 x 1024
• Training Batch Size: 1
• Gradient Accumulation Steps: 3
• Gradient Checkpointing: Enabled
• Learning Rate: 1e-4
• SNR Gamma: 5.0
• LR Scheduler: constant
• Warmup Steps: 0
• Precision: fp16 (mixed precision)
• Optimizer: 8-bit Adam
• Max Train Steps: 500
• Checkpointing Steps: 717
• Seed: 0