parallel LeafletFinder #47
Labels
Comments
|
Hello @orbeckst , I see that the base class parallelizes over the frames a trajectory has. Our parallel implementation of the LeafletFinder parallelized the execution over a single frame. I will use the most effecient case based on what we did for the paper, but I may have to override the core functionality of the based class and parallelize for a single frame and create multiple Dask DAGs to execute over the frames. Is that something you think will be usefull or do you prefer to keep the core functionality as is? |
|
The current framework might well be too limiting. I'm happy to consider other approaches!
It should present the same user interface but implementation details can differ.
…--
Oliver Beckstein
email: [email protected]
Am Aug 30, 2018 um 07:45 schrieb Ioannis Paraskevakos ***@***.***>:
Hello @orbeckst ,
I see that the base class parallelizes over the frames a trajectory has. Our parallel implementation of the LeafletFinder parallelized the execution over a single frame.
I will use the most effecient case based on what we did for the paper, but I may have to override the core functionality of the based class and parallelize for a single frame and create multiple Dask DAGs to execute over the frames.
Is that something you think will be usefull or do you prefer to keep the core functionality as is?
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub, or mute the thread.
|
|
@iparask can you link to the code that you used for the paper? |
|
Sorry for replying this late, but here is the code link. Please feel free to open tickets for discussion and any bugs that you may find if you want to run it. |
orbeckst
pushed a commit
that referenced
this issue
Oct 17, 2018
* fix #47 * NOTE: * follow-up in #72 (add PBC and other improvements) * This is the version of the algorithm published in the IPCP 2018 paper * Setting up parallel leaflet finder * Changed run method * Adding errors in LF and adding test * Adding Scipy and networkx in setup. Changed balltree with ckdtree * Single frame test passes, PEP8 both class and test. * Removing code comments based on PEP8 messages * trajectory level testing and pep8 changes * Changing where the paper is mentioned * Updated Changelog * Continuing PR review resolve * Reduced number of tasks * Removing redundant import * PEP8 corrections * Further PEP8 corrections * Debug, references, increased coverage and PEP8. * Docs added for Leaflet Finder * Addind missing file * Trying to fix linter's messages * Adding my name on AUTHORS, changelog and conf.py
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Parallel version of the
MDAnalysis.analysis.leaflet.LeafletFinderalgorithm from @iparask (see upcoming paper https://arxiv.org/abs/1801.07630).The text was updated successfully, but these errors were encountered: