lora-training

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

The main validation prompt used during training was:

A simple, clean line drawing of a right-angled triangle. The right angle is at the bottom left corner. The base is labeled as '6 cm' and the height is labeled as '4 cm'. The drawing is set against a plain white background.

Validation settings

CFG: 3.5
CFG Rescale: 0.0
Steps: 24
Sampler: None
Seed: 42
Resolution: 512

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, 3d, colorful

Prompt
A simple, clean line drawing of a right-angled triangle. The right angle is at the bottom left corner. The base is labeled as '6 cm' and the height is labeled as '4 cm'. The drawing is set against a plain white background.

Negative Prompt
blurry, cropped, ugly, 3d, colorful

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

Training settings

Training epochs: 8
Training steps: 3200
Learning rate: 0.0008
Effective batch size: 1
- Micro-batch size: 1
- Gradient accumulation steps: 1
- Number of GPUs: 1
Prediction type: flow-matching
Rescaled betas zero SNR: False
Optimizer: adamw_bf16
Precision: bf16
Quantised: Yes: int8-quanto
Xformers: Not used
LoRA Rank: 8
LoRA Alpha: None
LoRA Dropout: 0.1
LoRA initialisation style: default

Datasets

right-triangles

Repeats: 0
Total number of images: 380
Total number of aspect buckets: 1
Resolution: 512 px
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 = 'Mujeeb603/lora-training'
pipeline = DiffusionPipeline.from_pretrained(model_id)
pipeline.load_lora_weights(adapter_id)

prompt = "A simple, clean line drawing of a right-angled triangle. The right angle is at the bottom left corner. The base is labeled as '6 cm' and the height is labeled as '4 cm'. The drawing is set against a plain white background."

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=24,
    generator=torch.Generator(device='cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu').manual_seed(1641421826),
    width=512,
    height=512,
    guidance_scale=3.5,
).images[0]
image.save("output.png", format="PNG")