A Passionate Technologists' Journal

Fine-tuning TinyLlama Locally I recently fine-tuned TinyLlama on a small custom dataset and was impressed by how well it learned the specific response style. Here's what I did and the results. You can try it out yourself by checking out the repository . What is Fine-tuning? Fine-tuning takes a pre-trained language model (one that already understands general language) and trains it further on specific data to improve performance on particular tasks. Think of it as giving a general-purpose assistant specialized training in a specific domain. The Training Data I started with just 3 examples in a simple JSON format: [ {"prompt": "Explain Python lists", "response": "Python lists are ordered, mutable collections."}, {"prompt": "What is a dictionary?", "response": "A dictionary stores key-value pairs with fast lookup."}, {"prompt": "Explain list comprehension", ...

The Setup Picture this: you've got a CPU-intensive task—computing the sum of squares for 50 million numbers. You fire up Python with 4 threads, expecting to see your 4-core machine flex its muscles. But something's wrong. Your CPU usage hovers around 25%, and the program takes just as long (or longer) than if you'd used a single thread. What's happening? You've just met Python's Global Interpreter Lock (GIL), and it's not playing nice with your parallel dreams. The Experiment To understand this phenomenon, I built a simple benchmark: compute sum(i*i for i in range(N)) across multiple workers. This is pure CPU-bound work—no I/O, no waiting, just raw computation. The same task, implemented in three languages, tells a revealing story. Python (threads) : The GIL prevents multiple threads from executing Python bytecode simultaneously. Even with 4 threads, your program will mostly use 1 CPU core. You...

My Journey Exploring RAG vs GraphRAG My Journey Exploring RAG vs GraphRAG Introduction I knew what was RAG and GraphRAG, but didn't clearly understood how only the relationships in GraphRAG were performing better than simple/plain RAGs, so I did a POC to see if it does really make sense. There are many relationship based systems(fraud detection, social media recommendation, etc) where GraphRAG is really useful that could be picked up for the POC but I wanted something simple yet close to my passion, so I picked up the movie recommendation system which has entities/nodes like movies, actors, genres, directors, etc and relationship among them is easy to understand for everyone too. Tech Stack / Tools Neo4j for graph building logic TMDB API for movies data Open AI API for LLM integration ...

Introduction In the age of information overload, finding the most relevant moments in lengthy videos can be a daunting task. Clips Extractor is an innovative application designed to solve this problem by leveraging state-of-the-art Large Language Models (LLMs) and AI-powered transcription. This blog post explores the app's goals, technical architecture, and how it uses LLMs to deliver precise, topic-based video clips. What is Clips Extractor? Clips Extractor enables users to extract meaningful clips from YouTube videos based on a topic of interest. Whether you're a researcher, content creator, or casual viewer, you can quickly surface the most relevant segments without manually scrubbing through hours of footage. Key Features Extract clips from YouTube video Search for segments based on user-provided topics Get precise timestamps and transcripts for each clip Combine selected clips into a single video Chrome Extension for direct YouTube integration ...

I wrote a blog post in markdown format and tried formatting the codes here, but it felt like an endless task, so I decided to not continue with it, hence I'd request you to please read it out from the repo directly .