rochitasundar/Generative-AI-with-Large-Language-Models

Implements practical workflows for dialog summarization across three progressive stages: in-context learning with zero/few-shot prompting, instruction fine-tuning using LoRA and other PEFT methods on Flan-T5, and reinforcement learning with reward models (PPO) for toxicity reduction. Demonstrates evaluation techniques including ROUGE metrics and integrates Hugging Face models with parameter-efficient adaptation methods to reduce computational overhead while maintaining performance.

No commits in the last 6 months.