Apache Spark is a powerful platform for big data analytics that can now integrate seamlessly with deep learning models, thanks to recent developments. Utilizing NVIDIA RAPIDS and Spark APIs, it allows distributed training and inference even with large language models. Two main deployment strategies are explored: Basic inference
Table of contents
Why batch inference?Basic deployment: Distributed inference with predict_batch_udfAdvanced deployment: Distributed inference servingSummary: Choosing your deployment strategyDeploy on cloud platformsGetting startedSort: