npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2024 – Pkg Stats / Ryan Hefner

@aimr/asmr-gan-lib

v0.1.4

Published

![Npm publish passing](https://github.com/AI-ASMR/asmr-gan-core/actions/workflows/publish-npm-package.yml/badge.svg?branch=main) ![Git publish passing](https://github.com/AI-ASMR/asmr-gan-core/actions/workflows/publish-git-version.yml/badge.svg?branch=mai

Downloads

43

Readme

AiMR's Model Training

Npm publish passing Git publish passing Docker publish passing Npm package version GitHub release Codacy Badge Docker

Table of Contents

What is this?

This is a mono-repo hosting the source code of a GAN Neural Network model capable of producting (and being trained on) 64x64 images (1 or 3 channels), along with a helper/wrapper library that can be used in any javascript context to load a pre-trained model for immediate use.

Documentation can be found at ai-asmr.github.io/asmr-gan-core. Latest trained model files can be found here. The trained model files (model.json, weights.bin) are also uploaded to firebase and are publicly readable.

Model architecture.

Getting started.

The library can be found at npm and at the latest release. You can get it via npm like so.

npm i @aimr/asmr-gan-lib

Or you can install it locally by downloading the latest release.

npm i aimr-asmr-gan-lib-<version>.tgz

Or you can import it via cdn like so:

<script src="https://cdn.jsdelivr.net/npm/@aimr/asmr-gan-lib/.cdn/bundle.min.js"></script>

Example usage.

// or use require
import AiMR_GAN from "@aimr/asmr-gan-lib";

// import some version of tensorflow in any way you'd like
import * as tf from "@tensorflow/tfjs";

// load the model and/or cache it for subsequent reloads.
await AiMR_GAN.load(tf);

// generate some fake Asmr Images
console.log(await AiMR_GAN.generateChunks(1));

To train your own model take a look at the release binaries. To get a quick rundown on how to use it simply run:

asmr-gan-bin-<platform> --help

For better performance you can also use the binaries via docker like so:

# pull the latest version
sudo docker pull stiliyankushev/aimr-asmr-gan:latest
# run the docker instance (pass arguments at the end)
sudo docker run --gpus all -ti stiliyankushev/aimr-asmr-gan:latest --help

Full example of docker usage:

# assuming the training data is at "/home/kushev/Documents/training-data"
sudo docker run --mount src="/home/kushev/Documents",target="/home",type=bind \
--gpus all -ti stiliyankushev/aimr-asmr-gan:latest \
-i /home/training-data -d /home/dataset.bin -c /home/checkpoints -p /home/preview.png -q 10000 -s 512

(Optional) Docker Prerequisites.

Running the above docker container will automatically use a version of tensorflow that makes use of native C bindings. It'll also try to take advantage of any CUDA enabled GPUs running on the system. The docker container already pre-configures Cuda and Cudnn to work with tensorflow js. What you need to do is have:

  • Nvidia GPU with Cuda support.
  • Running a Linux distro.
  • Nvidia proprietary drivers installed.
  • Installed and configured NVIDIA Container Toolkit. (for arch linux, follow my guide.)

Build from source.

You can build both the library and the binary from source using short predefined npm-scripts. You need to install deps first.

git clone https://github.com/AI-ASMR/asmr-gan-core.git
cd ./asmr-gan-core
npm i

Build the library:

npm run build.lib

Build the binaries:

npm run build.bin

You can use the binaries directly from source without building executables. This will attempt to use your (CUDA enabled) GPU (Linux Only), same as with the docker container:

npm start -- --help

Requirements for CUDA enabled model training.

Running the above command will work but might not automatically pick up your GPU. That's why it's advised to use the docker image which comes pre-configured. However, if you'd like to run this locally without docker, here's what you need:

  • Nvidia GPU with Cuda support.
  • Running a Linux distro (preferably supported by tensorflow).
  • Cuda installed (version < v12.0.0).
  • Nvidia linux driver that supports the version of cuda installed.
  • libcudnn installed (version >= 8.9.5).

File structure.

This is a basic mono-repo that implements both a library and a running process. Both are their own separate typescript npm projects. Both share common assets/files, both share common npm packages listed in the root package.json and both extend root config files such as tsconfig.json.

+-- 📁 bin         # git release files here.
+-- 📁 common      # shared files between lib and src.
+-- 📁 lib         # sources of library here.
+-- 📁 src         # sources of binary here.
+-- 📁 tensorboard # tfjs storage used by bin.
+-- 📁 tests       # unit tests here.
+-- scripts.js     # mini build tool used by the repo.
+-- version.cfg    # version tracker.
+-- package.json
+-- README.md
+-- tsconfig.json

Versioning and automation.

Both the (npm) library and the (git) versioned binary, as well as the docker container share the same common version number. Versioning is automatically increased via CI/CD in the event of meaningful changes. Once there's a version change CI/CD automatically deploys updates for git releases, npm releases and docker tag releases respectively.

CI/CD implementation can be found here:

The repository hosts a minimal, scripted and cross-platform build tool used by all github actions, as well as users (via npm-scripts.)

For more details, read the documented source.

Arch linux NVIDIA Container Toolkit.

This is a short guide on how to install the NVIDIA Container Toolkit on arch linux. For other Linux distros take a look at their official guide.

I've created a custom PKGBUILD you need to build and install.

Make a fresh directory:

mkdir ./temp-nvidia 
cd ./temp-nvidia

Download the PKGBUILD file:

wget https://raw.githubusercontent.com/AI-ASMR/asmr-gan-core/main/PKGBUILD

Build the package:

makepkg

Install all .tar.zst files:

sudo pacman -U \ 
./libnvidia-container1-1.14.3-1-x86_64.pkg.tar.zst \
./libnvidia-container-tools-1.14.3-1-x86_64.pkg.tar.zst \
./nvidia-container-runtime-1.14.3-1-x86_64.pkg.tar.zst \
./nvidia-container-toolkit-1.14.3-1-x86_64.pkg.tar.zst \ 
./nvidia-container-toolkit-base-1.14.3-1-x86_64.pkg.tar.zst \ 
./nvidia-docker2-1.14.3-1-x86_64.pkg.tar.zst

Install libnvidia-container-tools manually:

sudo pacman -Syu libnvidia-container-tools

Configure docker:

sudo nvidia-ctk runtime configure --runtime=docker

Restart docker afterwards:

sudo systemctl restart docker

At this point docker should be configured. Test like so:

sudo docker run --gpus all ubuntu nvidia-smi

If nvidia-smi works, than everything works as expected.

Windows Support.

The easiest way to run this is to use the docker container in WSL(2) and enable NVIDIA CUDA following this guide.