Logo Lora

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

✨ Features

• Standard PEFT LoRA: Derived from the black-forest-labs/FLUX.1-dev base model.
• Specific Prompt Design: Trained with a specific prompt for generating car racing game app logos.
• Flexible Inference: Allows for reuse of the base model text encoder during inference.

📦 Installation

No specific installation steps are provided in the original document.

💻 Usage Examples

Basic Usage

import torch
from diffusers import DiffusionPipeline

model_id = 'black-forest-labs/FLUX.1-dev'
adapter_id = 'brianling16/logo-lora'
pipeline = DiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.bfloat16) # loading directly in bf16
pipeline.load_lora_weights(adapter_id)

prompt = "Generate a logo for a car racing game app"


## Optional: quantise the model to save on vram.
## Note: The model was quantised during training, and so it is recommended to do the same 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,
    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(42),
    width=512,
    height=512,
    guidance_scale=3.0,
).images[0]
image.save("output.png", format="PNG")

📚 Documentation

Training settings

• Training epochs: 0
• Training steps: 2600
• Learning rate: 8e-05
  • Learning rate schedule: polynomial
  • Warmup steps: 10
• Max grad norm: 2.0
• Effective batch size: 1
  • Micro-batch size: 1
  • Gradient accumulation steps: 1
  • Number of GPUs: 1
• Gradient checkpointing: True
• Prediction type: flow-matching (extra parameters=['shift=3', 'flux_guidance_mode=constant', 'flux_guidance_value=1.0', 'flow_matching_loss=compatible', 'flux_lora_target=all'])
• Optimizer: adamw_bf16
• Trainable parameter precision: Pure BF16
• Caption dropout probability: 5.0%
• LoRA Rank: 64
• LoRA Alpha: None
• LoRA Dropout: 0.1
• LoRA initialisation style: default

Datasets

app_data

• Repeats: 4
• Total number of images: 10001
• Total number of aspect buckets: 1
• Resolution: 0.262144 megapixels
• Cropped: False
• Crop style: None
• Crop aspect: None
• Used for regularisation data: No

🔧 Technical Details

The model is based on the black-forest-labs/FLUX.1-dev base model and uses the PEFT LoRA technique. During training, specific validation settings and training settings were applied, and the text encoder was not trained. The prediction type is flow-matching with specific extra parameters.

📄 License

The license is specified as other.