MBZUAI/BiMediX2-8B-hf
LoRA-style (fewer layers) vs meta-llama/Llama-3.1-8B-Instruct. meta-llama/Llama-3.1-8B-Instruct matches part of MBZUAI/BiMediX2-8B-hf (289 of its 686 tensors; the rest of MBZUAI/BiMediX2-8B-hf 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: Norms 0/65, MLP 96/96, Attention 128/128
2 derivatives checked
- mradermacher/BiMediX2-8B-hf-i1-GGUF: Quantization (fewer layers)
- mradermacher/BiMediX2-8B-hf-GGUF: Same weights, converted (fewer layers)