Laxhar/noobai-XL-0.75
Full fine-tune vs Laxhar/noobai-XL-0.6. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: Embeddings 590/590, other 662/662, Attention 716/716, Norms 546/546
3 derivatives checked
- Emanon14/NONAMEmix_v1: Full fine-tune
- John6666/cyberfalafel-falafel-cyber-fix-cyberfalafel-classic-sdxl: Different architecture
- Laxhar/noobai-XL-0.77: Full fine-tune
2 other repos hold byte-for-byte identical weight files.