pinkelephantlimited/pinkelephant-llm-48b-s
LoRA-style vs pinkelephantlimited/pinkelephant-llm-48b. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate, gate_up_proj changed, in 40 of 40 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/81, MLP 200/720, Attention 0/80, lm_head 0/1, other 0/2
1 derivatives checked
- pinkelephantlimited/pinkelephant-llm-48b-s-gguf: Not comparable yet