oriental-lab/Tr-Jp-LLM-1.5B-v2
Partial fine-tune vs SakanaAI/TinySwallow-1.5B-Instruct. Partial change: 20% of the weights changed (embeddings, mlp), in 6 of 28 layers.
Tensors changed: Embeddings 1/1, Norms 0/57, MLP 6/84, Attention 0/196
1 derivatives checked
- mradermacher/Tr-Jp-LLM-1.5B-v2-GGUF: Same weights, converted