LLM Fundamentals — Prompting & Text Generation Basics

Problem. Understand how large language models generate text and how to structure prompts for meaningful, controllable responses.

Approach.

• Introduced core LLM concepts: tokenization, context windows, temperature, top-k/top-p sampling.
• Demonstrated local or API-based text generation.
• Explored few-shot and zero-shot prompting patterns.
• Tested basic instruction tuning behavior using simple conversational inputs.
• Highlighted trade-offs between creativity, determinism, and coherence.

Results (qualitative).

• Identified how decoding parameters shape response variety.
• Observed improved contextual consistency with few-shot examples.
• Showcased reproducibility of responses under fixed random seeds.

What I Learned.

• Practical control of LLM inference parameters.
• Prompt design strategies for clarity and reliability.
• Relationship between token limits, context management, and output quality.

Quick Start

git clone https://github.com/Joe-Naz01/llm_basics.git
cd llm_basics

conda env create -f environment.yml
conda activate llm_basics

pip install -r requirements.txt
jupyter notebook