Trained on: https://www.deviantart.com/in-the-mind-of-ai 40 images
Trigger: aimindstyle
Works well with a combination of: sketch, digital media \(artwork)\, traditional media or any other quality tags.
Network Dimension Rank 128 - Network Alpha Rank 64
Learning Rates -> UNET 0.0005 - TE 0.00005
Optimiser -> AdamW8Bit, L2, Cosine with Restarts, 3 cycles
Batch Size 4, Epochs 10, Repeats 10
I'll be attaching the config toml and dataset too.
As for usage recommendations:
usual quality tags, adetailer (recommended but not always needed), hiresfix (Extra awesome)
works well with most models, some models look even better than the default illustrious give them a try too!
For further discussion on illustrious Lora's I have been documenting my journey here:
https://civitai.com/articles/9148