Tensara

<div align="center"> <picture> <img width="70%" alt="Tensara" src="https://github.com/user-attachments/assets/acd95491-2104-48d4-af84-bf2b23c95e72" /> </picture> </div>

Tensara is a platform for GPU programming challenges in CUDA, Triton, Mojo, etc. Users can write efficient GPU kernels to solve our problems and see how their solutions compare with others on the platform.

https://github.com/user-attachments/assets/96457139-2a27-493c-8352-df5ceb298369

Features

• Problems: Solve 60+ challenges in CUDA, Triton, and Mojo across multiple difficulty levels.
• Benchmarking: Run your solutions on actual GPUs (T4, H100, A100, etc.) with precise performance measurement.
• Leaderboards: Compare your performance against other developers on per-GPU rankings.
• Baseline Comparisons: See how your optimized kernels stack up against PyTorch, Triton, and other framework implementations
• CLI Tool: Submit and test solutions directly from your terminal with the Tensara CLI

Contributions

Sponsors

Thank you to our sponsors who help make Tensara possible:

• Modal - Modal lets you run jobs in the cloud, by just writing a few lines of Python. Customers use Modal to deploy Gen AI models at large scale, fine-tune large language models, run protein folding simulations, and much more.

We use Modal to securely run accurate benchmarks on various GPUs.

Contact

Interested in sponsoring? Contact us at sponsor@tensara.org or hit us up on Twitter!