Tag Archives: AI

MLOps roadmap 2024

Posted on by Leave a comment

The original content is in MLOps roadmap 2024, a wonderful article. I summarize the key points in the post for my reference.

IDDescriptionResources
1Programming
– Python & IDE
– Bash & command line editors		– IDE – visual studio code
– VIM
2Containerization and Kubernetes– Docker
– Kubernetes
3Machine learning fundamentalsa lot
4Machine learning fundamentalsBook: Introducing MLOps 𝖻𝗒 Mark Treveil 𝖺𝗇𝖽 Dataiku
5MLOps components– Git : Version control & CI/CD pipeline
– Airflow: Orchestration
– Mage-ai
– Mlflow: Experiment tracking and model registries
– Feast: feature store
– KubernetesPod Operator: Model training & serving
– nannyml: Monitoring & observability
– Evidently: Monitoring & observability
6Infrastructure as code– Terraform: Infrastructure as code

Moco based discriminative training

Posted on by Leave a comment

Moco based Fine-tune is good choice in duplicate/near duplicate tasks such as same image/ document detection, person identification, ….. amazing performance increase. As a discriminative training framework can fine-Tune any models vit, resnet, clip ….

Have fine-tuned resnet, vit, clip. AUC improvement is amazing.

Learn a metric oriented classifier

Posted on by Leave a comment

Objective function is the mathematical formulation of how to estimate classifier parameters. The classical objective function is derived from maximal log-likelihood function on training samples for the proposed classifier. Classifier parameters are estimated by solving the objective function. But log-likelihood is not directly related to performance metric, e.g. training on likelihood, and preferred evaluation metric maybe F1, accuracy or ranking. This criteria gap between training and evaluating causes the classifier trained on log-likelihood is not optimal for F1 , classification error or ranking. This is the intention of our work on MFoM based classifier learning. Updated the work on https://aisengtech.com/project#mfom. Hereafter MFoM, there are many research papers on learning classifier for specified metric in research community, in which learn-to-rank is most famous, and learn-to-rank is now a core module for modern search engine.