I tried phenix.maps again on a machine with more memory and it seems to work.  On my workstation (8GB) it eats 100% of memory and then 100% of swap and then dies.  I’d still be curious to get the official ensemble refinement map scripts to see if the result is the same but the phenix.maps map looks very reasonable.

Thanks,

Joel

From: Pavel Afonine [mailto:[email protected]]
Sent: Tuesday, July 23, 2013 3:16 PM
To: PHENIX user mailing list
Cc: Tom Burnley; Bard, Joel
Subject: Re: [phenixbb] calculate map from ensemble refinement

Hi Joel,

phenix.maps actually supposed to handle multi-model files provided that the occupancies are meaningful (equal to 1/N, where N is the number of models). But honestly I've never tried it with more than a few (5-10) models..

Pavel


On 7/23/13 8:20 AM, Tom Burnley wrote:

Hi Joel,

Short answer is no you need to run the jobs again.  However if they took an age to complete let me know as I have some scripts that can help reconstruct the maps.  Also please could you email me (off list) the input parameter file you used.  You are correct that the map_type=None likely caused the crash but I'd like to check that and make it so it intercepts this issue earlier in the code.

Cheers,

Tom

On 23 July 2013 17:05, Bard, Joel <[email protected]> wrote:

Hi-

I ran phenix.ensemble refinement which worked very nicely but crashed during map output.  I think this was my fault for using a parameter file that had map_type = None (I didn’t look and assumed the default was reasonable).  I would like to calculate maps after the fact.  Is there a way to run the map calculation on the output ensemble pdb file and reflection data?  I tried phenix.maps which choked on the huge ensemble.

Thanks,

Joel