Edward on Spark?


Does anyone know if edward can scale through Spark? Has any thought been given as to if this in on the roadmap?

Likewise, are there alternate tools that would be better suited to doing probablisitc programming on Spark if that is not Edwards main objective?




Edward has a lot of low-level flexibility, borrowing from TensorFlow and adding minimal abstractions on top. It’s compatible with the TensorFlow ecosystem, which includes Spark, Hadoop, Docker, Kubernetes, etc. The link has some documentation on how you can use Spark + TensorFlow. If there are issues specifically with Edward abstractions, please raise it on Github and we’ll be happy to solve it ASAP.