simpletuner-lora-flux-v2

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

The main validation prompt used during training was:

transparent objects on a table

Validation settings

CFG: 3.0
CFG Rescale: 0.0
Steps: 20
Sampler: None
Seed: 42
Resolution: 1024x1024

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
transparent objects on a table in low light

Negative Prompt
blurry, cropped, ugly

Prompt
transparent objects on a table in bright light

Negative Prompt
blurry, cropped, ugly

Prompt
transparent objects on a table in the backyard

Negative Prompt
blurry, cropped, ugly

Prompt
partially filled transaprent objects on a table

Negative Prompt
blurry, cropped, ugly

Prompt
transparent objects among opaque objects on a table

Negative Prompt
blurry, cropped, ugly

Prompt
transparent syringes on a table

Negative Prompt
blurry, cropped, ugly

Prompt
transparent objects in a container

Negative Prompt
blurry, cropped, ugly

Prompt
transparent objects on a table

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: 1
Training steps: 60000
Learning rate: 8e-05
Effective batch size: 4
- Micro-batch size: 1
- Gradient accumulation steps: 4
- Number of GPUs: 1
Prediction type: flow-matching
Rescaled betas zero SNR: False
Optimizer: adamw_bf16
Precision: bf16
Quantised: No
Xformers: Not used
LoRA Rank: 64
LoRA Alpha: None
LoRA Dropout: 0.1
LoRA initialisation style: default

Datasets

transparent_objects_custom

Repeats: 0
Total number of images: 160826
Total number of aspect buckets: 1
Resolution: 0.262144 megapixels
Cropped: False
Crop style: None
Crop aspect: None

Inference

import torch
from diffusers import DiffusionPipeline

model_id = 'black-forest-labs/FLUX.1-dev'
adapter_id = 'rohn132/simpletuner-lora-flux-v2'
pipeline = DiffusionPipeline.from_pretrained(model_id)
pipeline.load_lora_weights(adapter_id)

prompt = "transparent objects on a table"

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=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=1024,
    height=1024,
    guidance_scale=3.0,
).images[0]
image.save("output.png", format="PNG")