liswei/Taiwan-ELM-1_1B
Full fine-tune vs apple/OpenELM-1_1B. Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Norms 113/113, Attention 56/56, MLP 56/56
1 derivatives checked
- mradermacher/Taiwan-ELM-1_1B-GGUF: Same weights, converted