usermma/LFM2.5-230M-ECE
LoRA-style vs unsloth/LFM2.5-230M. 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 0/2, other 0/24, MLP 0/42, Norms 0/40, Attention 18/24
4 derivatives checked
- mradermacher/LFM2.5-230M-ECE-i1-GGUF: Quantization
- mradermacher/LFM2.5-230M-ECE-GGUF: Same weights, converted
- usermma/USELESS-Model-V1: LoRA-style
- usermma/LFM2.5-230M-ECE-OBLITERATED: Partial fine-tune