sd3-lora-training-rank128-v3

This is a standard PEFT LoRA derived from stabilityai/stable-diffusion-3-medium-diffusers.

The main validation prompt used during training was:

A design of a cute pokemon, featuring a single bird, water type pokemon, on a white background, style of Sugimori Kenji Vector Art

Validation settings

CFG: 5.0
CFG Rescale: 0.0
Steps: 20
Sampler: None
Seed: 42
Resolution: 1024x1024

Note: The validation settings are not necessarily the same as the training settings.

You can find some example images in the following gallery:

Prompt
unconditional (blank prompt)

Negative Prompt
blurry, cropped, ugly, missing body, missing limbs, missing ears, white noise

Prompt
A design of a cute pokemon, featuring a single bird, water type pokemon, on a white background, style of Sugimori Kenji Vector Art

Negative Prompt
blurry, cropped, ugly, missing body, missing limbs, missing ears, white noise

The text encoder was not trained. You may reuse the base model text encoder for inference.

Training settings

Training epochs: 59
Training steps: 4400
Learning rate: 0.0001
Effective batch size: 2
- Micro-batch size: 2
- Gradient accumulation steps: 1
- Number of GPUs: 1
Prediction type: flow-matching
Rescaled betas zero SNR: False
Optimizer: adamw_bf16
Precision: bf16
Quantised: No
Xformers: Not used
LoRA Rank: 128
LoRA Alpha: None
LoRA Dropout: 0.1
LoRA initialisation style: default

Datasets

pal_type

Repeats: 0
Total number of images: 73
Total number of aspect buckets: 1
Resolution: 1.048576 megapixels
Cropped: True
Crop style: center
Crop aspect: square

Inference

import torch
from diffusers import DiffusionPipeline

model_id = 'stabilityai/stable-diffusion-3-medium-diffusers'
adapter_id = 'zwloong/sd3-lora-training-rank128-v3'
pipeline = DiffusionPipeline.from_pretrained(model_id)
pipeline.load_lora_weights(adapter_id)

prompt = "A design of a cute pokemon, featuring a single bird, water type pokemon, on a white background, style of Sugimori Kenji Vector Art"
negative_prompt = 'blurry, cropped, ugly, missing body, missing limbs, missing ears, white noise'
pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')
image = pipeline(
    prompt=prompt,
    negative_prompt=negative_prompt,
    num_inference_steps=20,
    generator=torch.Generator(device='cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu').manual_seed(1641421826),
    width=1024,
    height=1024,
    guidance_scale=5.0,
).images[0]
image.save("output.png", format="PNG")