nvaitc-toolkit

KO | EN

gitlite — search

#python #java #python3 #arduino #golang #machine-learning #rust #html #flask #javascript #seismology #nodejs

★ 17 Open GitHub ↗

Open source code base to showcase interoperability of CUDA-X AI software stack in multi-GPU environments and thus provide researchers a reference framework to build new projects on.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

**NVIDIA AI Technology Centre (NVAITC)**
===============

# Toolkit

### Introduction

Python codebase to showcase the interoperability of CUDA-X AI software stack in multi-GPU environments. The goal of this project is to provide researchers a reference framework to build new projects on. It requests the availability of ImageNet to demonstrate how to train a network (ResNet[18/50/101]) against a well known dataset. This codebase served as the underlying playground for the Oct 2020 NVAITC Webinar Series on Deep Learning available as a [YouTube playlist](https://www.youtube.com/watch?v=f-CvtA2nuN8&list=PL5B692fm6--sJLzBmCpUSpP36xUWwuO8c).

### Clone repo

```
git clone -b toolkit --single-branch https://github.com/nvidia/nvaitc-toolkit.git toolkit
```

### Getting Started

Please find details and installation instructions in [README.md](https://github.com/NVIDIA/nvaitc-toolkit/blob/toolkit/README.md).

# cuAugment

### Introduction

cuAugment is a CUDA-accelerated 1D/2D/3D/4D augmenter library that utilizes a just-in-time compiler to transform a cascade of coordinate transformation into a single monolithic kernel to avoid unnecessary accesses to global memory.

### Clone repo

```
git clone -b cuaugment --single-branch https://github.com/nvidia/nvaitc-toolkit.git cuaugment
```

### Getting Started

Please find details and installation instructions in [README.md](https://github.com/NVIDIA/nvaitc-toolkit/blob/cuaugment/README.md).

# Build Docker container from scratch

The [Dockerfile](https://github.com/NVIDIA/nvaitc-toolkit/Dockerfile) available within this repository allows you build a new [docker](http://www.docker.com) container from scratch pulling both branches and installing packages needed to execute the code.

Before executing the docker build command please [download](https://developer.nvidia.com/nccl) the NCCL package from the NVIDIA Developer zone and edit the Dockerfile accordingly.

```
git clone https://github.com/nvidia/nvaitc-toolkit.git
cd nvaitc-toolkit
docker build .
```

nvaitc-toolkit

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README