From the course: Advanced Data Processing: Batch, Real-Time, and Cloud Architectures for AI
Join today to access over 24,500 courses taught by industry experts.
ML inference
From the course: Advanced Data Processing: Batch, Real-Time, and Cloud Architectures for AI
ML inference
- [Instructor] Having discussed feature engineering and model training, let's discuss ML inference in this video. Inference happens in production settings. Building an inference architecture for such a setting requires careful analysis of the tasks to be executed, the expected performance goals, and the infrastructure needed to achieve these goals. What are the tasks involved in model inference? First, raw data that is provided for inference need to be pre-processed and prepared for inference. In pre-processing, we need to ensure security of the model itself and protect it from intended and unintended hacks and misuse. Raw data may need to be temporarily stored or cached before it is processed. The feature engineering pipeline used for model training needs to be replicated to perform the same operations on inference data also. Reliable transformation of data needs to be insured either through APA calls or queues. Then comes serving. The model is usually in a model repository. The…
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.