Agora, alternative to Colab for AI/ML devs
Agora, alternative to Colab for AI/ML devs
Hi HN! We’re excited to introduce Agora, a GPU-backed Jupyter Notebook environment for AI/ML development. Agora aims to solve common problems with traditional solutions by: 1) sourcing the cheapest, high-performance compute in the world 2) creating a user-friendly way to use it for building in AI. Link: www.agoralabs.io Anyone can begin using Agora in three steps. 1) Create an Agora Lab with your custom GPU specs 2) Mount your Google Drive for storage/uploads 3) Launch JupyterLab instance to get to work! We created some cool features to make using Agora frictionless and compatible: - Ability to connect Google Drive & upload existing Colab notebooks, datasets, models, etc. - Source cheap, high-performance GPUs from decentralized networks like Vast.AI and Akash Network. - Built-in profiler so you know when you're getting close to running out of VRAM. We’re a team of three students; we tried to fine-tune Llama a few months ago & realized we needed GPUs. All of our work was on Jupyter Notebooks/Google Colab, but the dreaded OOM errors made them unusable. Paying for the Colab Pro version didn't get us access to high-performance compute like we needed, and renting GPUs from cloud services quickly became too expensive. We’ve seen a ton of AI/ML development start with notebooks (much of our own has), only to become stunted by compute and performance restrictions. We built Agora to solve our own problem and hopefully do the same for many others. People have been using Agora for sandbox-style experimentation work, like fine-tuning models for different tasks, testing various Agent frameworks, speeding up embeddings for personal notes, and more. With inspiration from Colab and Sagemaker, we built Agora with the close input of hackers and engineers. We welcome anyone interested in trying it out and would love to receive any feedback! [1] Link: https://agoralabs.io [2] Docs: https://docs.agoralabs.io/introduction
Share cardActual performance
Launch Intel predictions
Analyze your own launch →Incorrect prediction on native model
Similar products
Dockerfile Alternative for AI/ML
Lance – Alternative to Parquet for ML data
I created an alternative to Crowdfunding for devs, programmers etc.
Omegle for Devs
Shoot missiles at Ruby devs
Dumblr – tumblr for devs
I made a Crunchbase competitor using alternative data and ML
Brushless – An AI-first alternative to Illustrator
Signova AI – DocuSign alternative for $7
#1 Alternative to Supermetrics