Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

View file
nameAIMLFW-I-Release-Plan.pptx
height250

View file
name20230808-OSC Report.pdf
height250
View file
name20230808-Demo.mp4
height250

Topics:

  • H release maintenance release 
  • "I" Release plan discussion
    • "I" Release
    • RSAC presentation
    • EPICs added to JIRA
    • Jira tasks that can started in Dev sprint 1 to be added by next week.
    • Presentation and demo by NTUST

Discussion points:

  • Joseph Thaliath  mentioned that H release maintenance is ready
  • Professor and students from NTUST presented slides and showed on a demo on their implementation. They had some questions as mentioned below:
    • Q1. What is the feature store? What is inside it? Why called SDK?
      • Joseph Thaliath  clarified, that it is for storing features required for training. SDK is required to abstract the type of feature store DB and give a simple API for user.
    • Q2.How does the API update the job status and notify if the job status is changed, and what protocol is adopted?
      • Joseph Thaliath We have notification http post API which can be enabled. Other wise the current training job status can be queried using a GET API.
    • Q3. Can the general training pipeline do the training if the data source or characteristic is different?
      • Joseph Thaliath  clarified that the general training pipeline for training a particular types of AI/ML model (e.g. time series) and not specific to every usecase (e.g. qoe prediction) 
    • Q4. What is the purpose to have an automated testing for AI/ML Framework?
      • Joseph Thaliath clarified that we require an automated testing framework to test the installation of AIML framework and the APIs.
  • Rittwik Jana asked how a model is packaged with an rApp.
    • Joseph Thaliath clarified, that the model is deployed as separate pod by the kserve and it will expose http APIs to provide input and get prediction results.
  • Joseph Thaliath was proposing to have a deeper dive session with NTUST to discuss more on the collaboration.
  • Rittwik Jana asked if we have an idea about how to get data for the energy saving usecase. Joseph Thaliath said it is not clear now on how to get data.
  • Joseph Thaliath mentioned that EPICs for I release is created.
  • Joseph Thaliath mentioned that for generic training pipeline, we can have separate discussion with Navaneethan Rajasekaran 

2023-07-25 (Tuesday)

Recording Link: https://zoom.us/rec/share/03IX-dUJUDig4e2N1LcAqhJzbXYs8ysXuqlubwlHA6hzz5EOHnjf2CeY6jAsMX7c.lk_d6K_WGh3rEcnH

...