Text-to-image finetuning - iamkaikai/amazing-logos

This pipeline was finetuned from runwayml/stable-diffusion-v1-5 on the iamkaikai/amazing_logos_v2 dataset.

Training info

These are the key hyperparameters used during training:

• Dataset size: 10k
• Epochs: 20
• Learning rate: 1e-07
• Batch size: 1
• Gradient accumulation steps: 1
• Image resolution: 512
• Mixed-precision: fp16

Prompt Format

The prompt format is as follows:

{template keywords} + [company name] + [concept & country] + {template keywords}

For example:

Simple elegant logo for **[Google]**, **[G circle United states]**, successful vibe, minimalist, thought-provoking, abstract, recognizable, black and white

The [concept & country] section can include words such as:

• lines
• circles
• triangles
• dot
• crosses
• waves
• square
• letters (A-Z)
• 3D
• Angled
• Arrows
• cube
• Diamond
• Hexagon
• Loops
• outline
• ovals
• rectangle
• reflection
• rings
• round
• semicircle
• spiral
• woven
• stars

Here are some examples of prompts:

• Simple elegant logo for Digital Art, D A circle, successful vibe, minimalist, thought-provoking, abstract, recognizable, black and white
• Simple elegant logo for 3M Technology Products, 3 M square United states, successful vibe, minimalist, thought-provoking, abstract, recognizable, black and white
• Simple elegant logo for 38Energy, lines drop fire flame water, successful vibe, minimalist, thought provoking, abstract, recognizable, relatable, sharp, vector art, even edges, black and white