Skip to content

Latest commit

 

History

History
72 lines (59 loc) · 3.26 KB

meeting_00015_20201201.md

File metadata and controls

72 lines (59 loc) · 3.26 KB
Raw

Rust ML WG Meeting 00015

Meeting Info

Date: 20201201

Start time: 1600 ET

Zoom: https://ucsb.zoom.us/j/6601852842

Agenda

  • Discuss recent London meetup and talks
  • ndarray has been updated to v0.14
  • Release of linfa v0.2
    • Lots of new features now available through crates.io instead of just GH master
  • ONNX experiments
  • oxidize/enzyme

Participants

  • chrism
  • Tiberio
  • Ernest
  • Lorenz
  • Manuel
  • diversable (Daniel)

Minutes

  • Lorenz spoke with Luca and transferred ownership of linfa on crates.io to the Rust-ML group
    • version 0.2.1
  • Small announcement on Reddit about release on linfa
  • Going to speak with Paul re: ElasticNet implementation
  • Small issues with the appx_dbscan implementation, so may need to rebase
  • Lorenz is working on trees and porting over some tests
  • Great attendance at the London Rust community meetup (over 100 people at one point)
    • Highlighting corners of the Rust ecosystem that haven't necessarily been at the forefront (not just web servers)
    • Other talks have been about topics like GameDev
    • Got great feedback about Rust in data science. People didn't really know that this was an area people were using Rust, and Ernest heard a lot of great feedback and interest in the area
    • Quality of the hosting service wasn't great, esp. on audio
  • Rust London group is looking to add more talks later this year
    • If interested, reach out to Ernest on Zulip
  • diversable has expressed an interest in being able to interface Rust with Julia code
    • Julia has a lot of nice things: Mature-ish ML libraries, GPU support, accessible syntax
    • https://docs.rs/julia/0.2.5/julia/ offers a basic interface between the two
    • Possibly a new way to access consumers of Rust data science libs vs. producers
  • ndarray has been updated to 0.14, includes some new features and dependency bump, many more new updates on roadmap for 0.15
  • Rust-ML now has a twitter: @rust_ml
    • Good for community outreach
    • Ernest has found Rust interest in community groups in Rust Cambridge and Rust Oxford that may provide an audience that is interested in Rust data science and machine learning
      • Ernest would be happy to provide introductions
  • Discussion about Enyzme
  • ONNX
    • CPU-only, but used for creating portable computation graphs between neural network libraries
    • Being portable between different frameworks
    • Currently for inference, support for training is being worked on
    • Can be combined with ML-IR

Actions

  • Writing Book/examples using linfa 0.2
  • Docs improvements: adding a small example in each sub-crate to linfa to show how to use those algorithms
  • Continue experimenting with Enzyme for codegen
  • Discuss reaching out to Rust Oxford/Cambridge groups
  • If anyone has experience with Rust-Python bindgs, please reach out to Manuel, who is interested in moving forward with writing Python bindings for his Rust-RL project
  • Ernest may reach out with requests for more talks in the future, so anyone with interest can reach out him or keep that in the back of their mind