GitHub user mitstake edited a comment on the discussion: Parallelization Enhancement Ideas
Some potential issues I could see with this approach in Hamilton is that (1) it would be a departure from the “declarative” approach that Hamilton takes, as it’s pretty imperative in nature, but I don’t think that’s necessarily a bad thing. (2) It opens up alternative conflicting approaches to existing constructs in Hamilton, for example @config.when vs if/else statements before ngn.collect() to handle control flow. And (3) from a more technical point of view I believe that in Hamilton an instantiation of a driver is 1:1 with a defined graph, this would kind of blow that up since now a driver could have infinitely many graphs associated with it, which cannot be known until you execute a node and pass the parameters to it. Not sure how important these are to Hamilton, or that these are even an issue, but something to consider. GitHub link: https://github.com/apache/hamilton/discussions/1412#discussioncomment-15664572 ---- This is an automatically sent email for [email protected]. To unsubscribe, please send an email to: [email protected]
