franciscobdl/salamandra-estigiaV7
LoRA-style vs BSC-LT/salamandra-2b-instruct. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k_proj, o_proj, q_proj, v_proj changed, in 24 of 24 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/49, MLP 0/72, Attention 96/96
1 derivatives checked
- mradermacher/salamandra-estigiaV7-GGUF: Same weights, converted