Train a language model (e.g., Gemma-2B) using curated data from the kiali/kiali Hugging Face dataset, including common troubleshooting scenarios, Istio documentation, and UI usage patterns. Upload the resulting model to Hugging Face for versioned access.

Deliverables:

• Training scripts (using LoRA + 4-bit quantization)

• Dataset curation and annotation

• Upload and versioning in Hugging Face Model Hub