metadata

license: other
base_model: stabilityai/stable-diffusion-3.5-medium
tags:
  - sd3
  - sd3-diffusers
  - text-to-image
  - diffusers
  - simpletuner
  - not-for-all-audiences
  - lora
  - template:sd-lora
  - standard
inference: true
widget:
  - text: unconditional (blank prompt)
    parameters:
      negative_prompt: blurry, cropped, ugly
    output:
      url: ./assets/image_0_0.png
  - text: >-
      Baky tuxedo cat, resting on top of a table, looking to the camera, paws
      tucked
    parameters:
      negative_prompt: blurry, cropped, ugly
    output:
      url: ./assets/image_1_0.png

sd35-bucky_v7

This is a standard PEFT LoRA derived from stabilityai/stable-diffusion-3.5-medium.

The main validation prompt used during training was:

Baky tuxedo cat, resting on top of a table, looking to the camera, paws tucked

Validation settings

CFG: 7.5
CFG Rescale: 0.0
Steps: 10
Sampler: FlowMatchEulerDiscreteScheduler
Seed: 4201
Resolution: 1024
Skip-layer guidance:

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

You can find some example images in the following gallery:

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

Training settings

Training epochs: 5
Training steps: 640
Learning rate: 0.0002
- Learning rate schedule: polynomial
- Warmup steps: 64
Max grad norm: 2.0
Effective batch size: 2
- Micro-batch size: 2
- Gradient accumulation steps: 1
- Number of GPUs: 1
Gradient checkpointing: True
Prediction type: flow-matching (extra parameters=['shift=3'])
Optimizer: adamw_bf16
Trainable parameter precision: Pure BF16
Caption dropout probability: 0.0%
LoRA Rank: 768
LoRA Alpha: 768.0
LoRA Dropout: 0.1
LoRA initialisation style: default

Datasets

bucky-annotated-2

Repeats: 1
Total number of images: 120
Total number of aspect buckets: 2
Resolution: 1.0 megapixels
Cropped: False
Crop style: None
Crop aspect: None
Used for regularisation data: No

Inference

import torch
from diffusers import DiffusionPipeline

model_id = 'stabilityai/stable-diffusion-3.5-medium'
adapter_id = 'jwnt4/sd35-bucky_v7'
pipeline = DiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.bfloat16) # loading directly in bf16
pipeline.load_lora_weights(adapter_id)

prompt = "Baky tuxedo cat, resting on top of a table, looking to the camera, paws tucked"
negative_prompt = 'blurry, cropped, ugly'

## Optional: quantise the model to save on vram.
## Note: The model was not quantised during training, so it is not necessary to quantise it during inference time.
#from optimum.quanto import quantize, freeze, qint8
#quantize(pipeline.transformer, weights=qint8)
#freeze(pipeline.transformer)
    
pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu') # the pipeline is already in its target precision level
image = pipeline(
    prompt=prompt,
    negative_prompt=negative_prompt,
    num_inference_steps=10,
    generator=torch.Generator(device='cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu').manual_seed(4201),
    width=1024,
    height=1024,
    guidance_scale=7.5,
).images[0]
image.save("output.png", format="PNG")