M. Ayoub Chettouh Qfl3x

Week 2 of MLOps Zoomcamp

Link to Course: https://github.com/DataTalksClub/mlops-zoomcamp

Introduction to MLflow, ML experiments and model registry.

Introduction:

Definitions:

ML experiment: the process of building an ML model; The whole process in which a Data Scientist creates and optimizes a model

NOTE: commands and UI are deprecated

Content:

Negative Engineering
What is workflow orchestration?
Introduction to Prefect 2.0
First Prefect flow and Basics

Workflow Orchestration

Week 4 of MLOps-Zoomcamp

Link to Course: https://github.com/DataTalksClub/mlops-zoomcamp

Deployment

Quick recap:

We've learned how to rewrite our training into a workflow. Now we'll study how to deploy the resulting model.

In order to add a new IAM service account to publish on PubSub and download its key, we must:

Create a new IAM service account
Give it PubSub Editor role

Creating a new IAM service account:

To create a new IAM service account, simply follow the instructions provided here: https://cloud.google.com/iam/docs/creating-managing-service-accounts#creating

Giving it a PubSub Editor role:

Week 5 of MLOps Zoomcamp

Link to Course: https://github.com/DataTalksClub/mlops-zoomcamp

Monitoring

Why monitor ML models:

Our ML production models are production software and thus face the same problems faced by other production SE/SD software. However, in addition to these general issues, certain ML-specific issues may occur in ML production models that don't in SE/SD. As such, SE/SD tools are not sufficient to monitor ML production models.

Monitoring ML models is mostly around monitoring four sectors:

Best Practices

Testing with pytest:

We want to implement tests for the Lambda function.

Installing pytest as a dev dependency using pipenv:

pipenv install --dev pytest

	Computer Information:
	Manufacturer: Unknown
	Model: Unknown
	Form Factor: Desktop
	No Touch Input Detected

	Processor Information:
	CPU Vendor: GenuineIntel
	CPU Brand: Intel(R) Core(TM) i5-6500 CPU @ 3.20GHz
	CPU Family: 0x6

	GameAction [AppID 570, ActionID 5] : LaunchApp changed task to UpdatingAppInfo with ""
	GameAction [AppID 570, ActionID 5] : LaunchApp changed task to SynchronizingCloud with ""
	GameAction [AppID 570, ActionID 5] : LaunchApp changed task to ProcessingShaderCache with ""
	GameAction [AppID 570, ActionID 5] : LaunchApp changed task to SiteLicenseSeatCheckout with ""
	GameAction [AppID 570, ActionID 5] : LaunchApp changed task to CreatingProcess with ""
	GameAction [AppID 570, ActionID 5] : LaunchApp waiting for user response to CreatingProcess ""
	GameAction [AppID 570, ActionID 5] : LaunchApp continues with user response "CreatingProcess"
	Opted-in Controller Mask: 0
	Game update: AppID 570 "", ProcID 10977, IP 0.0.0.0:0
	ERROR: ld.so: object '/home/qflex/.local/share/Steam/ubuntu12_32/gameoverlayrenderer.so' from LD_PRELOAD cannot be preloaded (wrong ELF class: ELFCLASS32): ignored.