Lambent/Mira-v1.11-Ties-27B
Full fine-tune vs Lambent/Mira-1.10-dpo-27B. Fine-tune: 98% of the weights changed, across 62 of 62 layers.
Tensors changed: Embeddings 1/4, Norms 99/484, MLP 186/294, Attention 248/464, other 1/1
5 derivatives checked
- mradermacher/Mira-v1.11-Ties-27B-i1-GGUF: Quantization (fewer layers)
- mradermacher/Mira-v1.11-Ties-27B-GGUF: Quantization (fewer layers)
- Lambent/Mira-v1.12.1-27B: LoRA-style
- Lambent/Apeira-v0-27B: Full fine-tune
- Lambent/Unbound-v1.12.0-27B: LoRA-style