You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
One idea would be put all things into a yaml - the current input file, report, and the results of temperature/pressure sweep. We could also use a json for this, which would be faster for read in. For these quantities, I very much prefer a human readable format.
In addition, the raw backward/forward arrays that are integrated could be saved as npz format.
So I see that the ts results are already dumped into the temperature_sweep.dat Then I think a good first attack would be to add them to report.yaml and resave the raw tables from forward/backward.dat into npz's. Then in a second step we could think of removing the text files.
Do we want the full free_energies on each time/scaling step in the report or subsampled? I think like you, I'm slightly concerned that this would be a lot of data to lug around. Then again we are already writing it currently.
Activity
srmnitc commentedon Mar 11, 2026
One idea would be put all things into a yaml - the current input file, report, and the results of temperature/pressure sweep. We could also use a json for this, which would be faster for read in. For these quantities, I very much prefer a human readable format.
In addition, the raw backward/forward arrays that are integrated could be saved as npz format.
@pmrv do you have any thoughts?
pmrv commentedon Mar 12, 2026
Sounds good! I don't feel strongly about the format. Putting the raw data separately is nice, though. I can try to get started on this next week.
pmrv commentedon Mar 12, 2026
So I see that the ts results are already dumped into the temperature_sweep.dat Then I think a good first attack would be to add them to report.yaml and resave the raw tables from forward/backward.dat into npz's. Then in a second step we could think of removing the text files.
Do we want the full free_energies on each time/scaling step in the report or subsampled? I think like you, I'm slightly concerned that this would be a lot of data to lug around. Then again we are already writing it currently.
srmnitc commentedon Mar 13, 2026
I would be quite okay with subsampling per one K. We could dump the full arrays into the npz too.