OpenLLM-Ro/RoGemma-7b-Instruct-2024-10-09
Full fine-tune vs google/gemma-7b. Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 1/1, Norms 54/57, MLP 84/84, Attention 112/112
3 derivatives checked
- mradermacher/RoGemma-7b-Instruct-2024-10-09-GGUF: Same weights, converted
- OpenLLM-Ro/RoGemma-7b-Instruct-DPO: Full fine-tune
- OpenLLM-Ro/RoGemma-7b-Instruct-DPO-2024-10-09: Full fine-tune