Lambent/Mira-v1.12-Ties-27B
Full fine-tune vs Lambent/Mira-v1-dpo-27B. Fine-tune: 99% of the weights changed, across 62 of 62 layers.
Tensors changed: Embeddings 1/4, Norms 166/484, MLP 195/294, Attention 261/464, other 1/1
3 derivatives checked
- mradermacher/Mira-v1.12-Ties-27B-i1-GGUF: Quantization (fewer layers)
- mradermacher/Mira-v1.12-Ties-27B-GGUF: Quantization (fewer layers)
- Lambent/Mira-v1.13-27B-custom-heretic-ties: Full fine-tune