Ba2han/Karga-DPO-v0.1
LoRA-style vs Ba2han/Karga_EN-TR-8B-A1B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k_proj, q_proj, v_proj changed, in 6 of 24 layers; norms, biases and embeddings are untouched.
Tensors changed: MLP 0/2162, Embeddings 0/2, other 0/54, Norms 0/60, Attention 18/24
2 derivatives checked
- mradermacher/Karga-DPO-v0.1-i1-GGUF: Quantization
- mradermacher/Karga-DPO-v0.1-GGUF: Same weights, converted