rrw-x2/KoSOLAR-10.9B-v0.5
Full fine-tune vs LDCC/LDCC-SOLAR-10.7B. Fine-tune: 100% of the weights changed, across 48 of 48 layers.
Tensors changed: Embeddings 1/1, Norms 97/97, MLP 144/144, Attention 188/192, lm_head 1/1
1 derivatives checked
- mradermacher/KoSOLAR-10.9B-v0.5-i1-GGUF: Quantization
1 other repos hold byte-for-byte identical weight files.