vicgalle/SOLAR-13B-Instruct-v1.0
Partial fine-tune vs upstage/SOLAR-10.7B-Instruct-v1.0. Partial change: 41% of the weights changed (attention, mlp, norms), in 20 of 48 layers.
Tensors changed: Embeddings 0/1, Norms 40/97, MLP 60/144, Attention 80/192, lm_head 0/1
3 derivatives checked
- mradermacher/SOLAR-13B-Instruct-v1.0-i1-GGUF: Quantization
- mradermacher/SOLAR-13B-Instruct-v1.0-GGUF: Quantization
- vicgalle/solarized-13B-dpo: LoRA-style