Type
LORA
Publish Time
2024-06-16
Base Model
Pony
Version Introduction
Training run 41, epoch 000004
Dataset 16
Images
Same as in v2.0
Tagging
Same as in v2.0
Training
Script: https://github.com/bmaltais/kohya_ss commit 5ed5623
Dataset replicates/epoch: 10
Epochs: 4
Batch size: 2
Checkpoint: ponyDiffusionV6XL_v6StartWithThisOne.safetensors
Network configuration: LyCORIS/LoCon (dim=16, alpha=8, conv_dim=8, conv_alpha=1)
LR: Governed by Prodigy optimizer. Schedule was "cosine" annealed over 20 epochs, with 5% warmup.
Optimizer: Prodigy
Noise offset: Multires
CLIP skip: 2
