Changes between Initial Version and Version 1 of Ticket #17832, comment 3


Ignore:
Timestamp:
May 30, 2025, 11:53:12 AM (5 months ago)
Author:
Tom Goddard

Legend:

Unmodified
Added
Removed
Modified

  • Ticket #17832, comment 3

    initial v1  
    1 {{{
     1
    22Thanks for the response, Tom!
    33
    … …  
    1111Cheers,
    1212Alok
    13 ________________________________
    14 From: ChimeraX <ChimeraX-bugs-admin@cgl.ucsf.edu>
    15 Sent: Wednesday, May 28, 2025 7:27 PM
    16 To: goddard@cgl.ucsf.edu <goddard@cgl.ucsf.edu>; Alok Bharadwaj <A.Bharadwaj@tudelft.nl>
    17 Cc: chimerax-programmers@cgl.ucsf.edu <chimerax-programmers@cgl.ucsf.edu>
    18 Subject: Re: [ChimeraX] #17832: LocScale-SURFER bundle development for Mac OS
    1913
    20 [You don't often get email from chimerax-bugs-admin@cgl.ucsf.edu. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]
    21 
    22 #17832: LocScale-SURFER bundle development for Mac OS
    23 ------------------------------------+-------------------------
    24           Reporter:  A.Bharadwaj@…  |      Owner:  Tom Goddard
    25               Type:  defect         |     Status:  assigned
    26           Priority:  normal         |  Milestone:
    27          Component:  Tool Shed      |    Version:
    28         Resolution:                 |   Keywords:
    29         Blocked By:                 |   Blocking:
    30 Notify when closed:                 |   Platform:  all
    31            Project:  ChimeraX       |
    32 ------------------------------------+-------------------------
    33 Comment (by Tom Goddard):
    34 
    35  Hi Alok,
    36 
    37    I have never seen Conda packages installed into ChimeraX.  But the PyPi
    38  torch is version 2.7.0 so I think it is no different than the Conda torch.
    39  The difference is probably that you can tell torch to fallback to CPU
    40  computation for operations that are not available in MPS.  You do this by
    41  setting environment variable PYTORCH_ENABLE_MPS_FALLBACK = 1.  I use this
    42  in the ChimeraX Boltz structure prediction tool which also uses torch.  If
    43  your Conv3D layers are taking most of the time then probably MPS won't
    44  speed things up much if it has to fallback to CPU.  I think you can set
    45  this environment variable in Python.
    46 
    47      import os
    48      os.environ['PYTORCH_ENABLE_MPS_FALLBACK'] = '1'
    49 
    50  I am not sure if that will work.  In the ChimeraX Boltz tool I use pytorch
    51  in a subprocess rather than running directly in ChimeraX.
    52 
    53     Tom
    54 --
    55 Ticket URL: <https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.rbvi.ucsf.edu%2Ftrac%2FChimeraX%2Fticket%2F17832%23comment%3A2&data=05%7C02%7CA.Bharadwaj%40tudelft.nl%7Cae29e9b357af4e86fd8a08dd9e0cf07e%7C096e524d692940308cd38ab42de0887b%7C0%7C0%7C638840500639918935%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=ekC6iABOLClOG3ZbKzPykNW26UvCKiqAT%2F6%2B5mSuEbo%3D&reserved=0<https://www.rbvi.ucsf.edu/trac/ChimeraX/ticket/17832#comment:2>>
    56 ChimeraX <https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.rbvi.ucsf.edu%2Fchimerax%2F&data=05%7C02%7CA.Bharadwaj%40tudelft.nl%7Cae29e9b357af4e86fd8a08dd9e0cf07e%7C096e524d692940308cd38ab42de0887b%7C0%7C0%7C638840500639950869%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=7ESToj%2BUlw8U0VlglmAWXJBsEJf9UBHxqoxwjukHlW4%3D&reserved=0<https://www.rbvi.ucsf.edu/chimerax/>>
    57 ChimeraX Issue Tracker
    58 }}}