Skip to content

2019

Jump to bottom
Edgar Riba edited this page Mar 16, 2020 · 8 revisions

2019-11-20

Agenda

  • Discussion with @asmorkalov and @AnnaPetrovicheva from Xperience.ai about CI infrastructure.

Minutes

  1. Kornia uses Circle-CI as CI system right now.

    • There is one testing host and it's included into Cicle-CI plan.
    • The team wants to extend testing and include GPU.

  2. Releases are done by Edgar semi-automatically.

    • AI Edgar: send release instructions to Alexander for review.

  3. Amazon provides some AWS creadits for the project.

    • The team wants to do inital testing to understand price and available resources.
    • Alexander proposed test GPU workload on nightly basis to reduce cost, if it's required.
    • Pre-commit testing can be CPU only.

  4. Windows testing is painful.

    • PyTorch team runs dedicated Jenkins-CI instance and Windows hosts with GPU.
    • Windows support is not priority for the team.

  5. Docker is recommended solution for Linux as build and test environment.

    • NVIDIA provides NVIDIA Docker2 project to run containers with GPU support.

  6. Circle-CI and AWS related help and recommendations are welcome.

<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>

2019-10-31

Agenda

  • Organization
  • Community
  • Infrastructure
  • Roadmap

Minutes

  • Organization

  • Community

    • Number of PRs and contributors have increased a lot.
    • Make sure we have clear contribution protocols, tests. Revisit contributions docs.
    • Dmytro and Edgar will work on promotional slides (after cvpr deadline).
    • Benchmarks wrt to some popular functions
      • not a high priority now, just to have a brief idea. Include to slides.
    • Kaggle
      • Dmytro contact Jeremy to talk about kaggle kernels.

  • Infrastructure

    • Next release v0.1.5 December 1st.
      • release a new version every two months approx.
      • not broke compatibility, create guidelines.
    • Discuss with Camilo/RoboMake team from AWS about using CircleCI or CodeBuild for the GPU testing.
    • Anna will contact Alex to help with the GPU testing framework.
    • Conda packages:
      • not a priority now since we only depend on PyTorch.

  • Roadmap

    • See issue: https://github.com/kornia/kornia/issues/312
    • Data augmentation
    • GPU testing
      • prepare testing framework
      • setup CI
    • Epipolar geometry
    • Match opencv colorspaces conversions
    • Differenciable drawing
    • 3D reprojections/rendering
      • ping Vassilis from Scape

<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>

2019-09-04

Agenda

  • Roadmap
  • Organization
  • Infrastructure
  • Community

Minutes

  • Roadmap
    • GOAL: grow as community until Pytorch devcon (October 10th).
    • Push code used in the paper and do pre-release (v0.1.4).
    • Focus on improving documentation, examples and tutorials.
      • use Colab for examples and tutorials. Contact @priba, he did some examples already.
    • No new features in the short term. Plan new features roadmap after Pytorch devcon (v0.2).
  • Organization
    • @edgar @dmytro role as maintainers.
    • discussed with @gary about joining the OSVF.
      • @edgar will donate his gsoc part to kornia.
    • meetings schedule: monthly.
  • Infrastructure
    • @chadell will help with the devops.
    • mail server with google domains; @edgar owns kornia.org
    • Move to CircleCI; first CPU, next GPU.
      • Find CPU server for Discourse.
      • Find GPU server for testing; @gary has contacts with lambda; will talk to Ethan too.
  • Community
    • Discussion channel:
      • setup Discourse
      • CPU server 24h (need to look for options)
    • Website:
      • Should be ready before Pytorch devcon.
      • @carlosb1 proposeed Jekyll template in jekyll_kornia
      • will be hosted in github pages.
    • Promotion:
      • need to create slides
      • submit paper to arxiv
      • potential places to promote:
        • Pytorch devcon
        • @dmytro talk at his uni
        • OSVF
        • Pytorch ecosystem

To Dos

  • Edgar
    • contact @priba @diego to help out with examples
    • contact facebook to ask about resources for ecosystem projects.
    • ping the c++ developer and ask about status.
  • Gary
    • give @edgar contacts to access to gpu servers.
  • Luis
    • talk with Henry from Optima sports about the servers.

The Kornia.org team

Clone this wiki locally