KidIkaros/abliterated-minicpm5-2b
Partial fine-tune vs openbmb/MiniCPM5-2B. Partial change: 48% of the weights changed (attention, lm_head, mlp), in 20 of 42 layers.
Tensors changed: lm_head 1/1, Embeddings 0/1, Norms 0/85, MLP 60/126, Attention 80/168
3 derivatives checked
- mradermacher/abliterated-minicpm5-2b-i1-GGUF: Quantization
- mradermacher/abliterated-minicpm5-2b-GGUF: Same weights, converted
- KidIkaros/abliterated-minicpm5-2b-v2: LoRA-style