Fine-tuned model for professional interview question generation.
Model Info
Base: Qwen/Qwen2.5-3B-Instruct
Fine-tune: LoRA, merged
Task: Interview question generation (JSON output)
Max tokens recommended: 2200
Usage
Load directly with AutoModelForCausalLM.from_pretrained.
The LoRA was trained for professional interviewer tone/style. RAG should still provide company/job grounding, and the app should enforce JSON with prompt constraints plus validation/repair.
