Auraflow-DomoKun-LoRA-rank16

This is a standard PEFT LoRA derived from fal/AuraFlow-v0.2.

The main validation prompt used during training was:

An domokun running through a field with flowers all around him.

Validation settings

• CFG: 4.0
• CFG Rescale: 0.0
• Steps: 30
• Sampler: FlowMatchEulerDiscreteScheduler
• Seed: 42
• Resolution: 512x512

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

ugly, cropped, blurry, low-quality, mediocre average

Prompt

An domokun running through a field with flowers all around him.

Negative Prompt

ugly, cropped, blurry, low-quality, mediocre average

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

Training settings

• Training epochs: 0

• Training steps: 150

• Learning rate: 7e-05

  • Learning rate schedule: cosine
  • Warmup steps: 400000

• Max grad value: 0.0

• Effective batch size: 4

  • Micro-batch size: 4
  • Gradient accumulation steps: 1
  • Number of GPUs: 1

• Gradient checkpointing: True

• Prediction type: flow_matching (extra parameters=['flow_schedule_auto_shift', 'shift=0.0'])

• Optimizer: adamw_bf16

• Trainable parameter precision: Pure BF16

• Base model precision: no_change

• Caption dropout probability: 10.0%

• LoRA Rank: 16

• LoRA Alpha: 16.0

• LoRA Dropout: 0.1

• LoRA initialisation style: default

Datasets

domokun-uncropped-512

• Repeats: 10
• Total number of images: 27
• Total number of aspect buckets: 2
• Resolution: 0.262144 megapixels
• Cropped: False
• Crop style: None
• Crop aspect: None
• Used for regularisation data: No

domokun-cropped-512

• Repeats: 10
• Total number of images: 27
• Total number of aspect buckets: 7
• Resolution: 0.262144 megapixels
• Cropped: False
• Crop style: None
• Crop aspect: None
• Used for regularisation data: No

Inference

import torch
from diffusers import DiffusionPipeline

model_id = 'fal/AuraFlow-v0.2'
adapter_id = 'bghira/Auraflow-DomoKun-LoRA-rank16'
pipeline = DiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.bfloat16) # loading directly in bf16
pipeline.load_lora_weights(adapter_id)

prompt = "An domokun running through a field with flowers all around him."
negative_prompt = 'ugly, cropped, blurry, low-quality, mediocre average'

## 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
model_output = pipeline(
    prompt=prompt,
    negative_prompt=negative_prompt,
    num_inference_steps=30,
    generator=torch.Generator(device='cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu').manual_seed(42),
    width=512,
    height=512,
    guidance_scale=4.0,
).images[0]

model_output.save("output.png", format="PNG")

Exponential Moving Average (EMA)

SimpleTuner generates a safetensors variant of the EMA weights and a pt file.

The safetensors file is intended to be used for inference, and the pt file is for continuing finetuning.

The EMA model may provide a more well-rounded result, but typically will feel undertrained compared to the full model as it is a running decayed average of the model weights.