Danielbrdz/Barcenas-2x10.7b-Korean
LoRA-style (fewer layers) vs chihoonlee10/T3Q-ko-solar-dpo-v6.0. Danielbrdz/Barcenas-2x10.7b-Korean matches part of chihoonlee10/T3Q-ko-solar-dpo-v6.0 (291 of its 435 tensors; the rest of chihoonlee10/T3Q-ko-solar-dpo-v6.0 has no counterpart). On the shared part: Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 32 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/65, MLP 93/96, Attention 124/128, lm_head 0/1
2 derivatives checked
- mradermacher/Barcenas-2x10.7b-Korean-GGUF: Same weights, converted
- featherless-ai-quants/Danielbrdz-Barcenas-2x10.7b-Korean-GGUF: Quantization