Open rbouckaert opened 5 years ago
I mis-estimated the running time in a cluster batch submission system (is there at all a good way to get an estimate out of BEAST?), so I'd also be interested in resuming. Currently, NS does not store states at all, does it? But it seems to have inherited an incomplete implementation of resuming if there's accidentally a .xml.state
file around from an MCMC run.
Hi, is there any progress on this? If I understand correctly, current resume just starts new analysis that is logged to the same log files, instead of properly resuming.
Thanks!
From Kate at the BEAST user list: "I sadly don't have a UPS available for my desktop machine (hence the AWS time), which means that a power outage of less than ten seconds lost ten days of processing. I realise that it's probably a lot more complex to add this functionality than it looks, but it would be extremely reassuring to have it in place!"
What is required is to store/restore all active points + everything needed for the stopping criterion to be calculated.