psobot commented on pull request #13475: URL: https://github.com/apache/beam/pull/13475#issuecomment-754235246
Happy to include more details! One example use case: running TensorFlow inference in a Beam job, where TF can use multiple cores simultaneously when processing one element. Using a single-core VM restricts us to, well, one core - which means our processing happens slowly. Using a multi-core VM with only one SDK container allows us to use all of those cores from a single SDK container, by virtue of the fact that TensorFlow breaks out of the GIL and uses all available cores when processing an element. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
