mlabonne/FrankenMonarch-7B
Partial fine-tune (fewer layers) vs mlabonne/AlphaMonarch-7B. mlabonne/AlphaMonarch-7B matches part of mlabonne/FrankenMonarch-7B (291 of its 435 tensors; the rest of mlabonne/FrankenMonarch-7B has no counterpart). On the shared part: Partial change: 69% of the weights changed (attention, mlp, norms), in 23 of 32 layers.
Tensors changed: Embeddings 0/1, Norms 46/65, MLP 69/96, Attention 92/128, lm_head 0/1
1 derivatives checked
- mradermacher/FrankenMonarch-7B-GGUF: Quantization