Getting Started

TT-NN is a user-friendly API running ML workloads on Tenstorrent hardware.

The GitHub page for the project is located here: https://github.com/tenstorrent/tt-metal

Installation and environment setup instructions are in the installation guide.

1. Install and Build

Install and build the project by following the instructions in the installation guide.

2. Explore our model demos

Get started with the Falcon 7B demo. Navigate to the Falcon 7B demo folder for details.

You can also check our demos for ResNet, BERT, Mistral 7B, and Llama2-70B (coming soon on our T3000 platforms).

3. TT-NN Tutorial: Multi-Head Attention (Simple)

Learn the basics of multi-head attention operations with TT-NN with a simple example: TT-NN simple module.

4. TT-NN Tutorial: Multi-Head Attention (Optimized)

Dive deeper into multi-head attention operations in TT-NN, optimizing performance: optimizing performance.

Where to go from here

You should now take a look at more code examples for TT-NN or the other tutorials on using TT-NN with Jupyter Notebooks.

If you’re an internal TT-NN developer, please now read and review the contribution standards.