IMPORTANT ANNOUNCEMENT: Due to daylight savings in Europe (but not the US) the meetup is at 10am PST until the US decides to catch up to the rest of the world.
Also, I may have accidentally Rick Rolled you last week on the how to put ML into Prod link. Complete accident but now that so many ppl enjoyed it I may have to start doing it on purpose, consider yourself warned.
Our Medium
Speaking of getting models into production.....Part 2 of our series on putting models into production with SageMaker is now out! Community member Neylson Crepalde creator of the open source framework Hermione
and long time SageMaker Guru makes sure to break it down real slow and give us some helpful tips around Scalable SageMaker Endpoints and Batch Transform Jobs.
Side note, do you feel like writing for us? Write me back and let me know what you had in mind!
For our most recent Coffee Sessions, David and I sat down with Satish Chandra GuptaCo-founder of Slanglabs to talk about Data Engineering + ML + Software Engineering. He had an amazing blog post that I loved and we referenced many times in the chat around scalable and efficient big data analytics ML pipelines.
As you can see from the photo above he has much wisdom to impart on us all! i came up with like 3
different quotes from our chat that I will be putting out on our Twitter page over the next week! Check out the whole conversation in video or podcast form.
@Ilnardo92 akaIvan Nardini joined us last week to talk about Operationalizing Open Source Models with SAS Open Model Manager. I loved hearing him breakdown what so many of us already know about the different stakeholders in the ML process. Also he just wrote an amazing article on this subject for our medium page
which you can check out here.
In case you missed it the SAS Open Model Manager is a containerized modelops tool that accelerates deployment processes and, once in production, allows monitoring your models (SAS and Open Source). It was cool to see the demo and also talking with Ivan is always a pleasure!
KF serving is one of the most talked-about topics in the community so it only makes sense we get on someone who is overly excited about this stuff to do a deep dive workshop with us. Theofilos Papapanagiotou Data Science Architect at Prosus plans to give us a thorough walkthrough of kubeflows capabilities.
We will look inside some popular model formats like the SavedModel of Tensorflow, the Model Archiver of PyTorch, pickle&ONNX, to understand how the weights of the NN are saved there, the graph and the signature
concepts.
We will discuss the relevant resources of the deployment stack of Istio (the Ingress gateway, the sidecar and the virtual service) and Knative (the service and revisions), as well as Kubeflow and KFServing. Then we'll get into the design details of KFServing, its custom resources, the controller and webhooks, the logging and configuration.
Then we are going to spend a large part in the monitoring stack, the metrics of the servable (memory footprint, latency, number of requests), as well as the model metrics like the graph, init/restore latencies, the optimizations, and the runtime metrics which end up to Prometheus. We will look at the inference payload and prediction logging to observe drifts and trigger the retraining of the pipeline.
*Remember this week the timing is different if you
are in the Americas 10 am PST
Because you are doing great work and don't let anyone tell you any differently, unless you're working for facebook.
Have a great week! Check out our slack, youtube, and podcasts if you haven't already. Also, it would mean a lot to me if you filled out this form so I can learn more about the community.