📚 💰 Unleash ollama potential w. data and LLM capabilities

🕛 Zeit seit Veröffentlichung: 32 Tage, 7 Stunden 41 Minuten
📆 Veröffentlicht am: 03.05.2024 um 22:10 Uhr
💡 Newskategorie: Programmierung
🔗 Quelle: dev.to

💭 First thoughts

Since I discovered ollama many months ago, day after day the feeling that the number of models (and tags) was going ever bigger...somewhat out of control.

Lastly, news were so rich and models were landing at an ever increasing pace:

// Detect dark theme var iframe = document.getElementById('tweet-1780598146470379880-42'); if (document.body.className.includes('dark-theme')) { iframe.src = "https://platform.twitter.com/embed/Tweet.html?id=1780598146470379880&theme=dark" }

In the mean time, I was not the only one to find the were a lot of models out there :

// Detect dark theme var iframe = document.getElementById('tweet-1781431672262525094-130'); if (document.body.className.includes('dark-theme')) { iframe.src = "https://platform.twitter.com/embed/Tweet.html?id=1781431672262525094&theme=dark" }

👉 So I started to build a place to put all ollama library data altogether so anyone can use datascience on top of it for any purpose.

This post is all about this.

🎯 Goals

1. Get data together
2. Link data together
3. Schedule data delivery as Kaggle Dataset
4. Make some datastorytelling around MT-Bench and duckdb FTS (Full-Text Search Extension) and the automation opportunities we have

🍿 Demo

Below the full dataset unboxing

🤖 About automation, agentic, MT-Bench

🏆 Most trending models

🎨 Dataviz inspiration

// Detect dark theme var iframe = document.getElementById('tweet-1782799567689232764-194'); if (document.body.className.includes('dark-theme')) { iframe.src = "https://platform.twitter.com/embed/Tweet.html?id=1782799567689232764&theme=dark" }

...