vicgalle/franken-SOLAR-18B-v1.0
Partial fine-tune (fewer layers) vs NousResearch/Nous-Hermes-2-SOLAR-10.7B. NousResearch/Nous-Hermes-2-SOLAR-10.7B matches part of vicgalle/franken-SOLAR-18B-v1.0 (435 of its 732 tensors; the rest of vicgalle/franken-SOLAR-18B-v1.0 has no counterpart). On the shared part: Partial change: 73% of the weights changed (attention, mlp, norms), in 36 of 48 layers.
Tensors changed: Embeddings 0/1, Norms 72/97, MLP 108/144, Attention 144/192, lm_head 0/1
2 derivatives checked
- mradermacher/franken-SOLAR-18B-v1.0-i1-GGUF: Quantization
- vicgalle/solarized-18B-dpo: LoRA-style