Lambent/Mira-v1.25-27B-Wave
Full fine-tune vs Lambent/Mira-v1.21-27B-rlvr. Fine-tune: 100% of the weights changed, across 62 of 62 layers.
Tensors changed: Embeddings 4/4, Norms 481/484, MLP 294/294, Attention 460/464, other 1/1
4 derivatives checked
- mradermacher/Mira-v1.25-27B-Wave-i1-GGUF: Quantization (fewer layers)
- mradermacher/Mira-v1.25-27B-Wave-GGUF: Quantization (fewer layers)
- Lambent/Mira-v1.25.1-27B-DPO: Full fine-tune
- Lambent/Mira-v1.28-wave: Full fine-tune