Replies: 4 comments 19 replies
-
These again are probably intermediate warnings. As long as the system passes b.run_checks and the original forward model runs successfully, it shouldn't be an issue. How long does run_compute take? Contact binaries can be quite expensive, and each iteration in Nelder-Mead will require several calls to run_compute, so I'm not surprised if it takes quite a while. It's probably well worth the effort to try to optimize your forward model as much as possible first. This tutorial from the latest workshop covers some of the things you might want to try. |
Beta Was this translation helpful? Give feedback.
-
Hi Kyle,
I have now added in the code for the debugging logger, and although it made a file, it didn't write anything to it. Please find attached snapshots of my file so far. I have tried to keep it simple by commenting out some of the parameters I entered that I don't think are so important to see if that helped. And so far I only have one of my 3 passbands in the file. I hope the snapshots are enough for you to be able to see what I am trying to do. If you need anything else, please let me know.
Thanks again,
Kathryn
…________________________________
From: Kyle Conroy ***@***.***>
Sent: 28 July 2021 19:55
To: phoebe-project/phoebe2 ***@***.***>
Cc: kathmcp ***@***.***>; Author ***@***.***>
Subject: Re: [phoebe-project/phoebe2] Nelder-Mead optimizer (#508)
Ok... then I don't think speeding it up will likely fix that (although still will be useful to do once it is working). My guess is that the optimizer is wandering into an area of the parameter space where building the mesh is hanging.
Can you enable a debugging logger and send me the output? Immediately before run_solver, add phoebe.logger(flevel='DEBUG', filename='debug.log') (and remove any previous logger statements). Then just run the script, wait for it to stall again, and send me the last 50 or so lines of that log file.
If you can also send me a copy of the bundle immediately before calling run_solver, that would be easiest for me to try to reproduce. If you don't want to attach it publicly here, you can send it to me by email.
Thanks!
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub<https://emea01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fphoebe-project%2Fphoebe2%2Fdiscussions%2F508%23discussioncomment-1084089&data=04%7C01%7C%7Cc793396aabbc44c91d7608d95201b5e2%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637630989561188247%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=wbMYop6DVlfLadKlocmmcoQX6W3EdnaB6OuuP6iesbs%3D&reserved=0>, or unsubscribe<https://emea01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FATNTCQIHX6QWV6VZUARKE4DT2BOEVANCNFSM5BEWTPEA&data=04%7C01%7C%7Cc793396aabbc44c91d7608d95201b5e2%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637630989561188247%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=mKJkPTRZglXIJZtuPWaI682BNfzdMVfKuw0PB6T%2F1uA%3D&reserved=0>.
|
Beta Was this translation helpful? Give feedback.
-
Here is the csv file with the observations in it, if that's of any help.
Thanks,
Kathryn
…________________________________
From: Kyle Conroy ***@***.***>
Sent: 28 July 2021 19:55
To: phoebe-project/phoebe2 ***@***.***>
Cc: kathmcp ***@***.***>; Author ***@***.***>
Subject: Re: [phoebe-project/phoebe2] Nelder-Mead optimizer (#508)
Ok... then I don't think speeding it up will likely fix that (although still will be useful to do once it is working). My guess is that the optimizer is wandering into an area of the parameter space where building the mesh is hanging.
Can you enable a debugging logger and send me the output? Immediately before run_solver, add phoebe.logger(flevel='DEBUG', filename='debug.log') (and remove any previous logger statements). Then just run the script, wait for it to stall again, and send me the last 50 or so lines of that log file.
If you can also send me a copy of the bundle immediately before calling run_solver, that would be easiest for me to try to reproduce. If you don't want to attach it publicly here, you can send it to me by email.
Thanks!
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub<https://emea01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fphoebe-project%2Fphoebe2%2Fdiscussions%2F508%23discussioncomment-1084089&data=04%7C01%7C%7Cc793396aabbc44c91d7608d95201b5e2%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637630989561188247%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=wbMYop6DVlfLadKlocmmcoQX6W3EdnaB6OuuP6iesbs%3D&reserved=0>, or unsubscribe<https://emea01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FATNTCQIHX6QWV6VZUARKE4DT2BOEVANCNFSM5BEWTPEA&data=04%7C01%7C%7Cc793396aabbc44c91d7608d95201b5e2%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637630989561188247%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=mKJkPTRZglXIJZtuPWaI682BNfzdMVfKuw0PB6T%2F1uA%3D&reserved=0>.
|
Beta Was this translation helpful? Give feedback.
-
Hi Kyle, sorry, I thought I had attached the files in here last night, but they are not here, so here they are again, along with an xls file of the observations if you need it. thanks! |
Beta Was this translation helpful? Give feedback.
-
Hi Kyle,
I am now trying to run the Nelder-Mead optimizer, but it seems to be running very, very slowly. It is taking about 20s per iteration, and then after 18 iterations it seems to have just stopped. I removed the orb and mesh datasets, so it's only running one light curve, but this didn't seem to help. I also tried using fast compute with ellc, but it doesn't work for contact binaries. I am trying to just vary the mass ratio, initially.
When I first enter the parameters for my binary, I get the following errors and warnings (although it does run_compute later)
Wed, 28 Jul 2021 16:30 BUNDLE ERROR Constraint 'pot@contact_envelope@envelope@constraint' raised the following error while attempting to solve for 'pot@contact_envelope@envelope@component'. Consider flipping the constraint or changing the value of one of ['requiv@primary@star@component', 'q@binary@orbit@component', 'sma@binary@orbit@component', 'pot@contact_envelope@envelope@component'] until the constraint succeeds. Original error: roche_contact_Omega_at_partial_vol::Volume is outside bounds.
Wed, 28 Jul 2021 16:30 BUNDLE ERROR Constraint 'requiv@secondary@star@constraint' raised the following error while attempting to solve for 'requiv@secondary@star@component'. Consider flipping the constraint or changing the value of one of ['pot@contact_envelope@envelope@component', 'q@binary@orbit@component', 'sma@binary@orbit@component', 'requiv@secondary@star@component'] until the constraint succeeds. Original error: potential > L1 critical value
Wed, 28 Jul 2021 16:30 BUNDLE ERROR Constraint 'fillout_factor@contact_envelope@envelope@constraint' raised the following error while setting the value of 'fillout_factor@contact_envelope@envelope@component'. Original error: value of fillout_factor=-13.410026693369053 not within limits of [<Quantity 0.>, <Quantity 1.>]
Wed, 28 Jul 2021 16:30 BUNDLE WARNING constraints ['pot@contact_envelope@envelope@constraint'] failed to run. Address errors and try again. Call run_failed_constraints to see the tracebacks. If not addressed, this warning will continue to be raised and will throw an error at run_compute.
Wed, 28 Jul 2021 16:30 BUNDLE WARNING primary is underflowing at L1 and not a contact system (requiv=1.5, requiv_min=1.8346231503528594, requiv_max=1.9639267680718653) If not addressed, this warning will continue to be raised and will throw an error at run_compute.
Wed, 28 Jul 2021 16:30 BUNDLE WARNING secondary is overflowing at L2/L3 (requiv=1.4999999999999996, requiv_min=1.2725418568681297, requiv_max=0.9342870720587503) If not addressed, this warning will continue to be raised and will throw an error at run_compute.
Are these errors and warnings contributing to the problems with the optimizer?
Beta Was this translation helpful? Give feedback.
All reactions