Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • May 14 01:33
    poulson closed #276
  • May 14 01:33
    poulson commented #276
  • May 14 01:30

    poulson on master

    Update README.md (compare)

  • May 14 01:24

    poulson on master

    added logo (compare)

  • May 10 21:37
    jedbrown commented #276
  • May 10 21:25
    poulson commented #276
  • May 10 21:21
    jedbrown commented #276
  • May 10 17:08
    poulson commented #276
  • May 10 17:03
    tesch1 opened #276
  • Apr 10 13:18
    JM1 commented #275
  • Apr 10 13:16
    JM1 synchronize #275
  • Apr 10 08:56
    JM1 opened #275
  • Mar 06 03:47
    Raviteja1996 closed #274
  • Mar 05 05:46
    Raviteja1996 opened #274
  • Feb 11 21:53
    BenBrock commented #228
  • Feb 11 21:52
    BenBrock commented #228
  • Feb 11 21:51
    poulson commented #228
  • Feb 11 21:50
    poulson commented #228
  • Feb 11 21:45
    BenBrock commented #228
  • Jan 23 23:57
    adambaskerville commented #273
Ryan H. Lewis
@rhl-
Jack Poulson
@poulson
I can't request there
Ryan H. Lewis
@rhl-
why not?
you might need to sign up/log in with your FAS account
Jack Poulson
@poulson
ah; should I request builder?
Ryan H. Lewis
@rhl-
Not sure sounds good
Ryan H. Lewis
@rhl-
We could do ubsan within Jenkins I guess
Jack Poulson
@poulson
yes, it takes several times longer to compile, but that is okay to do every so often
it might be a bad idea for every commit
Ryan H. Lewis
@rhl-
Perhaps every time master advances
That's not too crazy
E.g. If the hash had advanced since the last build then rebuild
That should be close to what you want
Or just nightly
Jack Poulson
@poulson
I am going to remove the ModelFit and LogisticRegression implementations since they were never really finished
given that this seems to be the only way to keep people from using them
Ryan H. Lewis
@rhl-
Makes good sense
Ryan H. Lewis
@rhl-
I'm still struggling with BFGS, haven't had a chance to get back into it
Jack Poulson
@poulson
what are you struggling with?
Ryan H. Lewis
@rhl-
The line searches didn't seem to do what they needed to do. Eventually the hessian became indefinite. I need to be able to get some good unit tests around just the line search
Also, I am not coding this next week
Which doesn't help
Ryan H. Lewis
@rhl-
The line search seems to end up not wanting to accept steps no matter how small the get
Ryan H. Lewis
@rhl-
In order to give you access to elemental you have to become a fedora packager
Jack Poulson
@poulson
I can look into that tonight
thank you for letting me know
Jack Poulson
@poulson
I just merged the recent bug fixes into 0.87
I need to push out 0.87.6 and debian packages tonight
tom442
@tom442
Is it just me or is the ppa for xenial not working: I get "unable to locate package" ? cheers
Ryan H. Lewis
@rhl-
@tom442 when do you get that?
Ryan H. Lewis
@rhl-
ive just pushed elemental to fedora rawhide, 25, and 24.
if it all goes well then very soon yum install elemental* will install elemental :)
tom442
@tom442
after I do "sudo apt-get install elemental" even though both links indicated under "Technical details about this PPA" are in my /etc/apt/sources.list.
I guess the actual package name is "elemental-mpich-python2".
Ryan H. Lewis
@rhl-
Yeah that sounds right
I'm not sure if it's still there but look at elemental-all
It may exist or not
Ryan H. Lewis
@rhl-
@tom442 did that work out for you?
Jack Poulson
@poulson
it is elemental-mpich-python2
my apologies, the source package is named 'elemental'
I'm updating the documentation now
tom442
@tom442
Thank you for replying (and by the way thank you for this great library)
(Initially because I needed to compile some old code urgently (which worked like a charm on a version I had compiled on some old windows distro), I thought "great there's a ppa let's use that !")
"elemental-mpich-python2" installs without error messages, however my old code ran into trouble confirmed by test examples.
Therefore I built it from source (after apt-get remove "elemental-mpich-python2" !) also because I wasn't sure that the ppa would make the link with libflame correctly.
Ryan H. Lewis
@rhl-
The ppa is of more or less the most recent release which is more or less master
tom442
@tom442
I got segmentation faults when I compile code after installing Elemental with the ppa, which I don't get when I build it from source.
Jack Poulson
@poulson
libflame isn't needed for fast SVD anymore due to the recent Divide and Conquer SVD implementations
do you have an example driver that you can share that causes a segfault?
tom442
@tom442
Thanks for the information ! Just the simple:
#include <El.hpp>\n int main(){El::Initialize();El::Finalize(); return 0;}
caused a segfault! Also I remember I needed to include the path to "mpi.h" in the makefile which I don't need with the built from source version. So there was definitely some linkage problem for me. Sorry if I didn't track the bug more cautiously: I really have some urgent work to do and as the build version works well, I don't have the time to uninstall and retry the ppa now !
Ryan H. Lewis
@rhl-
use ` without spaces before and after code to format it :)
err.. three backticks
can you please show me how you compile
you can put it into something like fpaste.org if its long