BanglaLLM/BanglaLLama-3.2-3b-unlop-culturax-base-v0.0.3
LoRA-style vs meta-llama/Llama-3.2-3B. 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 28 of 28 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/57, MLP 84/84, Attention 112/112
2 derivatives checked
- mradermacher/BanglaLLama-3.2-3b-unlop-culturax-base-v0.0.3-GGUF: Same weights, converted
- BanglaLLM/BanglaLLama-3.2-3b-bangla-alpaca-orca-instruct-v0.0.1: Partial fine-tune