🚀 Welcome to Tenstorrent

Get started with AI inference and custom training on Tenstorrent hardware.

This extension provides interactive walkthroughs to help you set up your development environment, run inference on Tenstorrent devices, train custom models, and build production AI applications.

🔗 Quick Links

🛠️ Setup Information (Optional)

Already using Tenstorrent Cloud, Quietbox, or a preinstalled system? Your environment is ready - skip to the Learning Path below!

If you need to set up a fresh system, use tt-installer 2.0 for one-command installation:

Quick Install:

/bin/bash -c "$(curl -fsSL https://github.com/tenstorrent/tt-installer/releases/latest/download/install.sh)"

📖 Run Quick Install | 📥 Download Script | 📚 Full Documentation

tt-installer sets up drivers, firmware, tt-smi, HugePages, and tt-metalium containers in 5-15 minutes.

📚 Learning Path (39 Lessons)

Click on any lesson below to open it in the new interactive lesson viewer. Or click "Open Lessons Sidebar" above to browse all lessons in the sidebar.

Lessons are organized by category - start with "Your First Inference" for the quickest path to running models, or jump to "Custom Training" to train your own!

🚀 Your First Inference

  • 🔌Hardware Detection

    Scan for connected Tenstorrent devices and verify they're properly recognized by the system.

  • ✅Verify tt-metal Installation

    Test your tt-metal installation by running a sample operation on your Tenstorrent device.

  • 📥Download Model and Run Inference

    Download the Llama-3.1-8B-Instruct model and run inference on your Tenstorrent hardware.

  • 💬Interactive Chat with Direct API

    Build a custom chat application using tt-metal's Generator API directly.

  • 🌐HTTP API Server with Direct API

    Create a production-ready Flask API with the model loaded in memory.

🏭 Serving Models

  • 🖥️Production Inference with tt-inference-server

    Use Tenstorrent's official inference server for production deployments with simple CLI configuration.

  • 🏭Production Inference with vLLM

    Deploy with vLLM - OpenAI-compatible APIs, continuous batching, and enterprise features.

  • 🎨Image Generation with Stable Diffusion XL

    Generate high-resolution 1024x1024 images using Stable Diffusion XL Base on Tenstorrent hardware.

  • 🎬Video Generation with Stable Diffusion 3.5

    Create videos by generating frames with Stable Diffusion 3.5 - see hardware scaling from N150 to Galaxy!

🔧 Compilers & Tools

  • 🔨Image Classification with TT-Forge

    Explore TT-Forge - Tenstorrent's MLIR-based compiler! Start with validated models like MobileNetV2.

  • ⚡JAX Inference with TT-XLA

    Master TT-XLA - production-ready XLA compiler with simple wheel installation and multi-chip support.

🎯 Applications

  • 👨‍💻Coding Assistant with Prompt Engineering

    Build an AI coding assistant using Llama 3.1 8B and prompt engineering.

  • 🎞️Native Video Animation with AnimateDiff

    Learn to build standalone packages outside tt-metal! Create animated videos with AnimateDiff temporal attention.

🎓 Custom Training (8-Lesson Series)

  • 📖CT1: Understanding Custom Training

    Learn the fundamentals of custom training on Tenstorrent hardware. Understand fine-tuning vs training from scratch.

  • 📊CT2: Dataset Fundamentals

    Master dataset creation and validation for fine-tuning. Learn JSONL format, quality guidelines, and tokenization concepts.

  • ⚙️CT3: Configuration Patterns

    Learn YAML-driven training configuration using tt-blacksmith patterns. Master hyperparameters and device configuration.

  • 🔥CT4: Fine-tuning Basics

    Train a character-level language model from scratch on Tenstorrent hardware. Watch NanoGPT learn Shakespeare!

  • 🚀CT5: Multi-Device Training

    Scale training to multiple Tenstorrent chips with Data Parallel (DDP). Achieve 2-8x speedup on N300, T3K, and Galaxy.

  • 📈CT6: Experiment Tracking

    Master experiment tracking with file-based logging and Weights & Biases (WandB) integration. Make data-driven decisions!

  • 🏗️CT7: Architecture Basics

    Understand transformer architecture components before training from scratch. Learn embeddings, attention, and feed-forward networks.

  • 🌟CT8: Training from Scratch

    Build and train a transformer from random initialization. Design nano-trickster (11M params) and understand scaling laws!

👨‍🍳 Tenstorrent Cookbook

  • 📖Tenstorrent Cookbook Overview

    Welcome to the Tenstorrent Cookbook! Build 5 complete projects that teach fundamental TT-Metal techniques.

  • 🎮Recipe 1: Conway's Game of Life

    Build Conway's Game of Life using TTNN parallel tile computing. Learn convolution operations and cellular automata.

  • 🎵Recipe 2: Audio Signal Processing

    Build a real-time audio processing pipeline with TTNN. Compute mel-spectrograms and apply creative effects.

  • 🌀Recipe 3: Mandelbrot Fractal Explorer

    Render beautiful fractals with interactive zoom! Demonstrates GPU-style parallel computation.

  • 🖼️Recipe 4: Custom Image Filters

    Build a library of creative image filters using 2D convolution. From edge detection to artistic effects!

  • ⚛️Recipe 5: Particle Life Simulator

    Simulate emergent complexity from simple particle interactions! Features N² force calculations and multi-device acceleration.

🧠 CS Fundamentals

  • 🖥️Module 1: RISC-V & Computer Architecture

    Von Neumann architecture, fetch-decode-execute cycle, and RISC-V fundamentals.

  • 💾Module 2: The Memory Hierarchy

    Cache locality, bandwidth tradeoffs, and near-memory compute.

  • ⚡Module 3: Parallel Computing

    Amdahl's Law, SPMD patterns, and data parallelism. Scale from 1 to 880 cores!

  • 🌐Module 4: Networks and Communication

    Message passing, network topologies, and routing. Master the Network-on-Chip!

  • 🔒Module 5: Synchronization

    Race conditions, barriers, and coordination. Learn explicit synchronization!

  • 📚Module 6: Abstraction Layers

    From Python to machine code. Understand the compilation pipeline!

  • 📊Module 7: Computational Complexity in Practice

    Big-O meets real hardware. See why constants matter in algorithm-hardware co-design!

🎓 Advanced Topics

  • ⚙️Modern Setup with tt-installer 2.0

    The fastest way to get started! One-command installation of the full Tenstorrent stack.

  • 🤝Bounty Program: Model Bring-Up

    Learn how to contribute by bringing up new models. Master TT-Metal while becoming part of the ecosystem!

  • 🔍Exploring TT-Metalium

    Discover what's possible! Explore TTNN tutorials, browse the model zoo, and dive into programming examples.

  • ☁️Deploy tt-vscode-toolkit to Koyeb

    Deploy your own cloud-based VSCode IDE with the Tenstorrent extension pre-installed.

  • 🚀Deploy Your Work to Koyeb

    Deploy any Python application to Koyeb with Tenstorrent N300 hardware.

💡 Note: These lessons assume you have a working Tenstorrent environment. If you need to set up from scratch, see the "Setup Information" section above or start with the tt-installer lesson.

🛠️ Working Directory

This extension creates scripts and files in a dedicated workspace directory:

~/tt-scratchpad/

All generated Python scripts, configuration files, and test code will be saved here. This keeps your workspace organized and makes it easy to find and customize the code.

⚡ Quick Actions

📦 Create my ~/tt-scratchpad
📚 Open Lessons Sidebar
🔍 Detect Hardware
📖 Open Legacy Walkthrough

📖 Resources