autonomi-nos 0.0.1

Nitrous oxide system (NOS) for computer-vision.

nos 🔥: Nitrous Oxide System (NOS) for Computer Vision

NOS is a PyTorch library for optimizing and running lightning-fast inference of popular computer vision models. NOS inherits its name from "Nitrous Oxide System", the performance-enhancing system typically used in racing cars. NOS is designed to be modular and easy to extend.

Why NOS?

• ⚡️ Fast: NOS is built on top of PyTorch and is designed to run models faster.
• 🔥 Out-of-the-box Performance: Run stable diffusion or object detection in under 5 lines, 2-3x faster than vanilla PyTorch.
• 👩‍💻 Reduce barrier-to-entry: NOS is designed to be easy to use. No ML optimizations or compilers knowledge necessary.
• 📦 Modular: NOS is designed to be modular and easy to extend. Optimize Pytorch models in a few lines of code.
• ⚙️ HW-accelerated: NOS is designed to leverage hardware-acceleration down to the metal (GPUs, TPUs, ASICs etc).
• ☁️ Cloud-agnostic: NOS is designed to run on any cloud (AWS, GCP, Azure, Lambda Labs, on-prem, etc.).

Batteries Included

• 💪 SOTA Model Support: NOS comes with support for popular CV models such as Stable Diffusion, ViT, CLIP, and more.
• 🐳 Docker: NOS comes with optimized docker images for accelerated CV workloads (runtime libs, drivers, optimized models).
• 🔌 Interfaces: NOS comes with a REST/gRPC API out-of-the-box to help you use your models.
• 📈 Benchmarks: NOS comes with a suite of benchmarks to help you compare performance of your models.

Hardware Support

We currently plan to support the following hardware:

• GPU (NVIDIA GPUs, AMD GPUs)
  • AWS (g4/g5dn/p3/p4dn)
  • GCP (g2/a1/n1)
• AWS Inferentia inf1/inf2
• Intel Habana Gaudi
• Google TPUv3

Lint

make lint  # Runs all the linters using ruff/pre-commit

Test

make test  # Runs all the basic tests using pytest

Benchmark

make test-benchmarks  # Runs all the benchmarks setting NOS_TEST_BENCHMARK=1

Contribute

We welcome contributions! Please see our contributing guide for more information.