mamelles/LFM2.5-350M-Wolof-CPT-v3
LoRA-style vs LiquidAI/LFM2.5-350M-Base. 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 16 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, other 0/30, MLP 0/48, Norms 0/44, Attention 18/24