mamelles/LFM2.5-230M-Wolof-CPT-v3
LoRA-style vs LiquidAI/LFM2.5-230M-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 14 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, other 0/24, MLP 0/42, Norms 0/40, Attention 18/24