Nvidia-MTC

Compile fast ML models for NVIDIA GPUs.

Attribution

• TVM - ML powered ML compiler.
• TCLPack - Prebuilt TVM binaries for NVIDIA CUDA
• Modal - Serverless GPU FAAS platform.

Prerequisites

You will need a Modal account. The app should fit within free tier.

1. Create a .env file in project root with the following contents:

APP_NAME=some-random-app-name
MODAL_WORKSPACE=your-modal-workspace