Hi,
> Does the same hold for training?
Training needs more memory.
"number_of_particles_in_a_movie" becomes
"number of particle used for training" (typically a few thousands).
I myself don't run training in most cases, except for very
important datasets.
Best regards,
Takanori Nakane
On 2019/09/22 10:48, Wim Hagen wrote:
> Hi,
>
> Does the same hold for training?
>
> Best,
>
> Wim Hagen
>
>> On 18. Sep 2019, at 13:46, Takanori Nakane <[log in to unmask]> wrote:
>>
>> Hi,
>>
>> Roughly speaking, we need the sum of the total movie size,
>> the extracted particle size and some working area, which is much
>> smaller compared to the other two.
>>
>> The total movie size is movie_width * movie_height * frame * 4 (float).
>> The extracted particle size is number_of_particles_in_a_movie *
>> box_width_in_movie * box_height_in_movie * frame * 4.
>>
>> In typical (non super-resolution) datasets, this is up to 20 GB.
>> So I usually run 2 MPI processes per 64 GB node, but it is
>> safer to run only 1 process per node if you have hundreds of
>> particles with overlapping boxes.
>>
>> Best regards,
>>
>> Takanori Nakane
>>
>>> Dear all,
>>>
>>> How does one calculate memory needed for Bayesian polishing?
>>>
>>> Best,
>>>
>>> Wim Hagen
>>> EMBL Heidelberg
>>>
>>> ########################################################################
>>>
>>> To unsubscribe from the CCPEM list, click the following link:
>>> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
>>>
>>
>> ########################################################################
>>
>> To unsubscribe from the CCPEM list, click the following link:
>> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
>
> ########################################################################
>
> To unsubscribe from the CCPEM list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
>
########################################################################
To unsubscribe from the CCPEM list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
|