sacred Execution of experiments in queue

Hell, Thanks for a very useful tool! After some doc exploration and usage I've still got 2 related questions:

If there is an out of the box possibility to execute queued experiments?
if it's possile to re-run already finished experiment considering we've got all the necessary dependecies and resources?

Cheers, Evgeny

Mar 03 '17 11:03 Evgeny7777

Hi Evgeny,

unfortunately the answer to both is "not yet". In an earlier version of the current release I had a prototype Assistant that could do an early version of both. Due to a lack of time to polish and document it, however, I removed it again. If you are interested you can find it here

I agree that both these features would be very useful, and the first one is even the reason the query option exists in the first place. I just haven't gotten around to properly implementing them yet. The above assistant still needs access to the experiment object, but in principle even that could be fetched from the database. This will likely be part of the next release, but I cannot promise a concrete date.

Mar 06 '17 17:03 Qwlouse

Any updates? I am willing to work on this feature but I'd rather not start from scratch.

Jul 12 '23 23:07 gwf

Hi @gwf!

No, there are unfortunately no updates on this. I would be happy to accept a PR for this feature but the link posted by @Qwlouse is the best/only starting point right now.

A related issue: https://github.com/IDSIA/sacred/issues/483. It might be a good idea to start by implementing some interface to fetch information about experiments stored by different observers (Mongo, file storage, ...). But I guess this is a lot of work...

Jul 13 '23 06:07 thequilo

Hey @thequilo,

Thanks for the pointers. I am going to try to get something working for Mongo first (as that's what I use and can easily test).

I'll reply here if/when I have something worth sharing.

Jul 13 '23 21:07 gwf