Activity of NVIDIA/TensorRT repository

Active 🚀

Active contributions

Why NVIDIA/TensorRT is active?

The result is based on ratio of number of commits and code additions from initial and final time ranges.

Initial time range – from 5 Jul, 2023 to 5 Oct, 2023

Final time range – from 5 Apr, 2024 to 5 Jul, 2024

From 0 to 1 commits per week

Infinity%

From 7 to 1430 additions per week

20329%

From 2 to 1060 deletions per week

52900%

Data calculated on 5 Jul, 2024

Bus factor

What is Bus factor?

It is basically a number of most active contributors responsible for 80% of contributions.

Bus factor tries to assess "What happens if a key member of the team is hit by a bus?". The more there are key members, the lower the risk.

The NVIDIA/TensorRT repository has a bus factor of 14.

Low risk, knowledge is well distributed among the team members

Bus factor was measured on 14 Aug 2024

Summary of NVIDIA/TensorRT

The NVIDIA/TensorRT GitHub repository houses TensorRT, NVIDIA's deep learning optimization library and runtime for production deployment. TensorRT is used to optimize deep learning models, providing impressive speedups and making it applicable for deployment in fields like self-driving cars and robotics where real-time execution speed is crucial.

Here's a quick overview of key components:

High-performance Inference Optimizer: TensorRT uses an optimizer to take trained models from any major framework and, using a large kernel library, optimize them to run efficiently on NVIDIA GPUs.
Runtime: Once the deep learning models have been optimized, the runtime is used to execute them. The runtime is accessible from both C++ and Python, providing implementation flexibility.
Automatic Mixed Precision: TensorRT features automatic mixed-precision capabilities, allowing users to leverage the Tensor Cores on NVIDIA's latest GPUs for even faster inference times.
Compatibility with Deep Learning Libraries: TensorFlow, PyTorch, MXNet, ONNX, and other open-source frameworks are supported.
Custom Layers and Plugins: For rare operations that aren't already supported by TensorRT's extensive library, you can add your own through custom layers and plugins.

In the repository, you'll find a thorough README file, as well as multiple folders containing samples, tools, and docs to assist in understanding and utilizing TensorRT's capabilities.

Here is the link to the repository: NVIDIA/TensorRT

To get started with TensorRT, users can refer to the detailed documentation and instructions available in the repository. Also, users can raise issues and contribute to the repository which are essential aspects of the open-source community.