grundprinzip commented on PR #40695: URL: https://github.com/apache/spark/pull/40695#issuecomment-1501388453
> I assume you mean to run torch program in spark connect client machine, we can support this, but I think it is less meaningful, because client machine should usually run lightweight workloads, but torch programs are heavy workloads and they often requires GPU, which client machine are hard to satisfy the condition. I don't think this is a valid assumption. With Spark Connect you can actually build an environment in which you have GPU locally but don't have a GPU on your cluster. In this case you still want to leverage the same execution flow. I've previously talked to users that were looking for an EC2 setup with a GPU attached and running the workloads from there against Sprk using Spark Connect. This is very similar to running sklearn locally on the client side. It's not the only way, but it's a very valid way. -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
