It is excellent news that we can now have a larger log file, and download 35GB! This should make a huge difference to my ability to tune the model. I'm interested to know more about how this will work in practice. At the moment, I have a job running. Will it have the new 1GB log file limit - or will that only be active in the future? How can we access the 35GB in the trained model? Is there somewhere where I can find instructions on how to do this?

Created by Peter Brooks fustbariclation
Hi Peter, Documentation for trained model retrieval is on its way. The feature will be enabled first on the Express Lane before the opening of the competitive phase. We are still fine tuning the specifications, but currently your method can place model state(s) in a 1 GB partition (/modelState). When a submission ends (for a reason or another), the content of /modelState will be archived on Synapse and returned to you (email + link in the dashboard to download it).

Trained Model Retrieval Update - - your Newsletter on the 7th of November page is loading…