flux-training-losercity-next

This is a LoRA derived from black-forest-labs/FLUX.1-dev.

Example prompts:

prompts = [
    'In this scene from the animated series "Helluva Boss," Loona, the wolf-like receptionist of the Immediate Murder Professionals (I.M.P), is depicted leaning against a wall outside the office. She is casually engrossed in her phone, displaying her typical aloof and detached demeanor. Loona\'s appearance includes her usual whitish fur, light grey hair, black-tipped ears, and red eyes, complemented by her punk-inspired attire featuring a black choker with spikes, a dark grey top, fingerless wrist-length black gloves, and black shorts.',
    'Loona shrugs with an exasperated expression, her red eyes wide and frustrated, as she seemingly questions or challenges something said in the I.M.P office. Still from Helluva boss. Loona\'s appearance includes her usual whitish fur, light grey hair, black-tipped ears, and red eyes, complemented by her punk-inspired attire featuring a black choker with spikes, a dark grey top, fingerless wrist-length black gloves, and black shorts.',
    "A scene from the animated series \"Helluva Boss,\" set in the office. Loona, the wolf-like receptionist with white fur, black-tipped ears, and red eyes, is seated on a couch, facing towards the viewer. Loona\'s appearance is complemented by her punk-inspired attire featuring a black choker with spikes, a dark grey top, fingerless wrist-length black gloves, and black shorts. She holds a piece of paper that says,\"Welcome to Losercity, jerks\". In the background, the office has a striped wall pattern and visible damage on the ceiling, indicating a chaotic or rough environment. On the right side of the image, two imp characters appear to be engaged in conversation.",
    "Loona from Helluva Boss is dressed in an oversized taco costume, looking visibly irritated and embarrassed. Her red eyes convey her annoyance as she crosses her arms and glares to the side. Loona\'s appearance includes her usual whitish fur, light grey hair, black-tipped ears, and red eyes",
]

To use Loona in classic style, just add the following trigger sentence to your prompt:

Loona's appearance includes her usual whitish fur, light grey hair, black-tipped ears, and red eyes, complemented by her punk-inspired attire featuring a black choker with spikes, a dark grey top, fingerless wrist-length black gloves, and black shorts.

The main validation prompt used during training was:

loona from helluva boss is eating a donut

Validation settings

• CFG: 3.5
• CFG Rescale: 0.0
• Steps: 15
• Sampler: None
• Seed: 42
• Resolution: 1024

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

Prompt

loona from helluva boss is eating a donut

Negative Prompt

blurry, cropped, ugly

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

Training settings

• Training epochs: 428
• Training steps: 3000
• Learning rate: 0.0001
• Effective batch size: 6
  • Micro-batch size: 6
  • Gradient accumulation steps: 1
  • Number of GPUs: 1
• Prediction type: flow-matching
• Rescaled betas zero SNR: False
• Optimizer: AdamW, stochastic bf16
• Precision: Pure BF16
• Xformers: Enabled
• LoRA Rank: 64
• LoRA Alpha: None
• LoRA Dropout: 0.1
• LoRA initialisation style: default

Datasets

losercity

• Repeats: 0
• Total number of images: 42
• Total number of aspect buckets: 1
• Resolution: 1.0 megapixels
• Cropped: True
• Crop style: center
• Crop aspect: square

Inference

import torch
from diffusers import DiffusionPipeline

model_id = 'black-forest-labs/FLUX.1-dev'
adapter_id = 'jimmycarter/flux-training-losercity-next'
pipeline = DiffusionPipeline.from_pretrained(model_id)
pipeline.load_lora_weights(adapter_id)

prompt = "loona from helluva boss is eating a donut"


pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')
image = pipeline(
    prompt=prompt,
    num_inference_steps=15,
    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=3.5,
).images[0]
image.save("output.png", format="PNG")