Restart from intermediate step?
Hi,
I'm just getting started with RapidMiner (which seems very cool and useful), so my first question is probably a simple, maybe even naive, one:
I have a operator chain in which I've created a neural net model. I made some changes downstream from the model creation step, and wanted to re-execute the chain. However, the neural net takes about 90 min to generate. But since the changes are all downstream, I was wondering if RapidMiner knows that the results from previous steps are not affected, and would be able to pick up processing from where the first change is? Either that, or can I tell RapidMiner to start from a particular operator, without reevaluating all the previous steps?
Is this RapidMiner's behavior, and if not, is there some way to get to what I described?
Thanks,
Keith
I'm just getting started with RapidMiner (which seems very cool and useful), so my first question is probably a simple, maybe even naive, one:
I have a operator chain in which I've created a neural net model. I made some changes downstream from the model creation step, and wanted to re-execute the chain. However, the neural net takes about 90 min to generate. But since the changes are all downstream, I was wondering if RapidMiner knows that the results from previous steps are not affected, and would be able to pick up processing from where the first change is? Either that, or can I tell RapidMiner to start from a particular operator, without reevaluating all the previous steps?
Is this RapidMiner's behavior, and if not, is there some way to get to what I described?
Thanks,
Keith
Find more posts tagged with
Sort by:
1 - 1 of
11
If you do not consider "saving intermediate results to disc using operators" and "reloading in a separate process" as an option, then no. If the RapidMiner were able to do something like this, it would either save large tempfiles to disc (what you can do manually) or keep it in the memory (which you do not want, believe me
So...using an Operator like "ModelWriter" is approriate here.
hope this was helpful
Steffen