跳至主要内容

Infrastructure & Tooling for Julialang Ecosystem

The Big Picture​

infra&tool

Taken from Full Stack Deep Learning Spring 2021

The figure mainly shows the infrastructure and tools used in the Python ecosystem, I will follow the same concept to collect the infrastructure and tools in the julia ecosystem.

Data​

Sources​

Filesystem​

  • locally mounted disk
  • networked (e.g. NFS)

Database​

  • PostgresQL
  • SQLite
  • InfluxDB: for time series
  • QuestDB: for time series

Versioning​

  • DVC

For more information on data version control in Julia, please refer to this discussion thread

Exploration​

  • DataFrames.jl

Training/Evaluation​

Computing​

  • desktop
  • workstation

Resource Management​

  • Docker

Software Engineering​

  • Julia v1.6.5, LTS release
  • VSCode
  • Jupyter Notebook
  • Pluto Notebook
  • git

Frameworks & Distributed Training​

Experiment Management​

Hyperparameter Tuning​

See use case for using Hyperopt.jl

Deployment​

CI/Testing​

  • GitHub Action

Edge​

  • ONNX.jl

Appendix​

Julia Package​

Useful Utility​

  • Revise
  • Test
  • ReTest
  • PkgTemplates
  • BenchmarkTools
  • Chain: piping
  • LoggingExtras: Composable Loggers for the Julia Logging StdLib
  • Memento: A flexible logging library for Julia
  • JuliaFormatter
  • DotEnv or ConfigEnv: loads environment variables from a .env file into ENV
  • WandbMacros or WeightsAndBiasLogger or Wandb: logging to weights and biases (Wandb) dashboard.

Database​

  • LibPQ: LibPQ.jl is a Julia wrapper for the PostgreSQL libpq C library.

ML​

  • Flux: framework
  • GeometricFlux: Geometric Deep Learning for Flux
  • GraphNeuralNetworks
  • FluxArchitectures: Complex neural network examples for Flux.jl.
  • Transformer: Julia Implementation of Transformer models
  • Metalhead: Computer vision models for Flux
  • MLJ: ML framework

ML Preprocessing​

  • Augmentor: A fast image augmentation library in Julia for machine learning.
  • MLDataUtils.jl
  • MLUtils.jl

Multi-threading/Multi-processing/Distributed​

  • Dagger: A framework for out-of-core and parallel execution
  • Polyester

Math/Statistic​

  • GeoStats: An extensible framework for high-performance geostatistics in Julia.
  • QuadGK
  • Random
  • Distributions
  • LinearAlgebra
  • StatsBase
  • Statistics
  • DSP

Data table and manipulation​

  • DataFrames: In-memory tabular data in Julia
  • Tables: An interface for tables in Julia
  • FeatureTransforms: Transformations for performing feature engineering in machine learning applications
  • TableTransforms: Transforms and pipelines with tabular data
  • Impute: Imputation methods for missing data in julia
  • DataConvenience: Convenience functions missing in Julia
  • Cleaner: A toolbox of simple solutions for common data cleaning problems.

Data IO​

  • Arrow
  • Parquet
  • CSV
  • JSON3
  • BSON

Notebook​

  • IJulia
  • Pluto

Visualization​

  • Plots
    • Backends
      • GR
      • PlotlyJS, PlotlyBase
      • PyPlot
    • Extensions
      • StatsPlots
      • GraphRecipes
  • Makie
    • Backends
      • CairoMakie
      • GLMakie(need GPU)
      • WGLMakie
    • Extensions
      • AlgebraOfGraphics
      • GraphMakie
      • GeoMakie
  • VegaLite
  • Gadfly
  • Compose
  • Colors
  • ColorSchemes

Text processing​

  • LaTeXStrings

Build in packages (standard library)​

  • LinearAlgebra
  • Statistics
  • Markdown
  • Printf