https://github.com/elefthei/webtorch

A Deep Learning HTTP server based on nginx
https://github.com/elefthei/webtorch

Last synced: 4 months ago
JSON representation

A Deep Learning HTTP server based on nginx

• Host: GitHub
• URL: https://github.com/elefthei/webtorch
• Owner: elefthei
• License: gpl-3.0
• Created: 2017-09-05T02:42:14.000Z (almost 9 years ago)
• Default Branch: master
• Last Pushed: 2018-09-10T18:04:40.000Z (almost 8 years ago)
• Last Synced: 2025-07-22T14:46:48.490Z (12 months ago)
• Language: Lua
• Homepage:
• Size: 83 KB
• Stars: 5
• Watchers: 1
• Forks: 2
• Open Issues: 3
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
[![Build Status](https://travis-ci.org/elefthei/WebTorch.svg?branch=master)](https://travis-ci.org/elefthei/WebTorch) [![contributions welcome](https://img.shields.io/badge/contributions-welcome-brightgreen.svg?style=flat)](https://github.com/elefthei/WebTorch/issues)

![WebTorch logo](https://elefthei.github.io/assets/css/images/webtorch.jpg)

WebTorch takes advantage of the versatile LuaJIT compiler to bring together Nginx and Torch, creating an equivalent system to Tensorflow serve, but based on a RESTful API instead of protobuf. Multiple data formats are supported (tensors, images etc.) 

## Building WebTorch container

Building (docker/nvidia-docker):

```

docker build -t webtorch .

```

## Running WebTorch container

Running daemon(docker/nvidia-docker):

```

docker run -p 3000:3000 -d webtorch

```

Running in debug(docker/nvidia-docker):

```

docker run -p 3000:3000 -it webtorch

```

## Usage:

### GET /v1/train

Will train the XOR neural network with a single random (input1, input2) sample.

```

$ curl http://localhost:3000/v1/train

```

### POST /v1/train

*Body*: 2D serialized Torch tensor (see `benchmarks/script/train.lua` for example)

Will train the XOR neural network with a serialized 2D Torch tensor sample.

```

$ wrk -t12 -c400 -d30s -s benchmarks/scripts/train.lua http://localhost:3000/v1/train

```

### GET /v1/output

Will get the result of the trained XOR network on (0.5, 0.5)

```

$ curl http://localhost:3000/v1/output

```