Harry Potter Multi-Persona QA Model V2 with 60K QA SFT (LoRA Adapter)
This is a LoRA adapter for google/gemma-2b fine-tuned for Harry Potter universe question-answering with multiple character personas.
Supported Personas - Hermione Granger: Analytical, articulate, and deeply knowledgeable - Harry Potter: Courageous, emotional, and direct - Severus Snape: Cold, precise, and cutting - Voldemort: Grandiose, manipulative, and authoritative - David Goggins: Raw, gritty, and brutally honest - Donald Trump: Boastful, blunt, and repetitive - General: Expert on Harry Potter universe
Training Details - Base Model: google/gemma-2b - Training Method: Supervised Fine-Tuning (SFT) with LoRA - Dataset: Custom Harry Potter QA dataset with persona-specific responses - Max Sequence Length: 512 tokens - Batch Size: 4 (with gradient accumulation)
Model Details - Developed by: Jayesh Bankoti - Model type: Causal Language Model with LoRA adapter - Language(s): English - Finetuned from model: google/gemma-2b