liswei/Taiwan-ELM-270M
Full fine-tune vs apple/OpenELM-270M. Fine-tune: 100% of the weights changed, across 16 of 16 layers.
Tensors changed: Norms 65/65, Attention 32/32, MLP 32/32
1 derivatives checked
- mradermacher/Taiwan-ELM-270M-GGUF: Same weights, converted