Training epochs: 9
Training steps: 8770
Learning rate: 6e-06
Max grad norm: 0.01
Effective batch size: 3
- Micro-batch size: 3
- Gradient accumulation steps: 1
- Number of GPUs: 1
Prediction type: flow-matching (extra parameters=['shift=3'])
Optimizer: adamw_bf16
Trainable parameter precision: Pure BF16
Quantised base model: No
Xformers: Not used
LoRA Rank: 64
LoRA Alpha: 64.0
LoRA Dropout: 0.1
LoRA initialisation style: default

Datasets

BESCH-CROP-SD35L-V02-512

Repeats: 1
Total number of images: 104
Total number of aspect buckets: 1
Resolution: 0.262144 megapixels
Cropped: True
Crop style: random
Crop aspect: closest
Used for regularisation data: No

BESCH-CROP-SD35L-V02-768

Repeats: 1
Total number of images: 104
Total number of aspect buckets: 1
Resolution: 0.589824 megapixels
Cropped: True
Crop style: random
Crop aspect: closest
Used for regularisation data: No

BESCH-CROP-SD35L-V02-1024

Repeats: 1
Total number of images: 104
Total number of aspect buckets: 3
Resolution: 1.048576 megapixels
Cropped: True
Crop style: random
Crop aspect: closest
Used for regularisation data: No

BESCH-CROP-SD35L-V02-1280

Repeats: 1
Total number of images: 104
Total number of aspect buckets: 3
Resolution: 1.6384 megapixels
Cropped: True
Crop style: random
Crop aspect: closest
Used for regularisation data: No

BESCH-MIX-SD35L-V02-512

Repeats: 1
Total number of images: 215
Total number of aspect buckets: 3
Resolution: 0.262144 megapixels
Cropped: True
Crop style: random
Crop aspect: closest
Used for regularisation data: No

BESCH-MIX-SD35L-V02-768

Repeats: 1
Total number of images: 215
Total number of aspect buckets: 2
Resolution: 0.589824 megapixels
Cropped: True
Crop style: random
Crop aspect: closest
Used for regularisation data: No

BESCH-MIX-SD35L-V02-1024

Repeats: 1
Total number of images: 214
Total number of aspect buckets: 3
Resolution: 1.048576 megapixels
Cropped: True
Crop style: random
Crop aspect: closest
Used for regularisation data: No

BESCH-MIX-SD35L-V02-1280

Repeats: 1
Total number of images: 212
Total number of aspect buckets: 1
Resolution: 1.6384 megapixels
Cropped: True
Crop style: random
Crop aspect: closest
Used for regularisation data: No

Inference

python

import torch
from diffusers import DiffusionPipeline

model_id = 'stabilityai/stable-diffusion-3.5-large'
adapter_id = 'gattaplayer/besch-style-st-sd35l-lora-6e-6-bs3-v02'
pipeline = DiffusionPipeline.from_pretrained(model_id), torch_dtype=torch.bfloat16) # loading directly in bf16
pipeline.load_lora_weights(adapter_id)

prompt = "An astronaut is riding a horse through the jungles of Thailand."
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=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=832,
    height=1216,
    guidance_scale=4.5,
).images[0]
image.save("output.png", format="PNG")