FENE problems in version HOOMD/0.11.3

Teresa Ruiz Herrero

unread,

Sep 24, 2014, 4:40:08 PM9/24/14

to hoomd...@googlegroups.com

Hi,

I have been running different old versions of hoomd without problems, now I am trying to move to a newer version so I can use its new properties and I am experiencing problems with the FENE bonds.

HOOMD/0.10.1 and previous versions run without issues my configurations and my previous scripts, however when trying to run the same initial configurations and scripts with HOOMD/0.11.3 I get the following errror at the very beginning:

** starting run **
**ERROR**: bond.fene: bond out of bounds

I realized that the bond syntax is changed, but even using the new one I get this error.

Anyone has encountered this problem before?

Thanks

Teresa

Jens Glaser

unread,

Sep 24, 2014, 5:07:49 PM9/24/14

to hoomd...@googlegroups.com

Teresa,

versions 0.10 and 0.11.3 are outdated… have you tried with 1.0.1 yet?

Jens

--
You received this message because you are subscribed to the Google Groups "hoomd-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hoomd-users...@googlegroups.com.
To post to this group, send email to hoomd...@googlegroups.com.
Visit this group at http://groups.google.com/group/hoomd-users.
For more options, visit https://groups.google.com/d/optout.

Teresa Ruiz Herrero

unread,

Sep 24, 2014, 5:23:38 PM9/24/14

to hoomd...@googlegroups.com

no I haven't tried, these are the versions installed in the clusters I am using, do you think that error was due to any bug on version 0.11?

thanks

Teresa

Ryan Marson

unread,

Sep 24, 2014, 5:30:13 PM9/24/14

to hoomd...@googlegroups.com

There was a bug in the FENE code that Jens corrected at some point over the past few months. I remember triggering it, though I don't recall the details of the bug. It could very well have been corrected between the two versions.

I would definitely try it with a newer version of the code!

-Ryan

Jens Glaser

unread,

Sep 24, 2014, 5:34:00 PM9/24/14

to hoomd...@googlegroups.com

Teresa,

According to https://codeblue.umich.edu/index/projects/hoomd-blue/wiki/ChangeLog, there has been one bond-related bug fix in 1.0:

Fixed a bug in bond.* where random crashes would occur when more than one bond type was defined

and one FENE related in 0.11.0 (I am not sure if that causes the change in behavior you are seeing)

Fixed a bug where FENE bonds were sometimes computed incorrectly

If you have no means of trying out 1.0.1 on your cluster, you can send me a script to reproduce your bug and I can try it with a more recent version

Since the bond framework has undergone a major (internal) overhaul in 1.0, I highly recommend using the latest version.

Jens

Teresa Ruiz Herrero

unread,

Sep 24, 2014, 6:22:28 PM9/24/14

to hoomd...@googlegroups.com

Hi Jens,

I would be very thankful if you try my code with the latest version, if it works I could ask to get that version installed in the cluster.
I attach here the configuration file and the script.

Thanks

Teresa

mpr.hoomd

init0.xml

Teresa Ruiz Herrero

unread,

Sep 28, 2014, 9:59:20 PM9/28/14

to hoomd...@googlegroups.com

Hi,

I managed to get hoomd1.0.1 compiled and tried to run my scripts with my configurations. WIth the latest version I experience the same problem I found with hoomd 0.11, the system blows up almost immediately after initialization, if I remove the fene bonds, there is no problem. I also tried with configurations obtained from my own MD code, again they only work with the previous hoomd versions.

So I am still facing the same problem, for hoomd versions older than hoomd0.10 everything works fine with my configurations created both with old hoomd or with my own MD codes, with newest hoomd versions however , fene bonds don't work correctly.

I wondered whether the system intitialization for new versions could be different, but I don't find in the documentation any essential change.

Is it possible there is any bug in the code for fene bonds?

Thanks

Teresa

the system inmediately blows up

Ryan Marson

unread,

Sep 29, 2014, 11:33:15 AM9/29/14

to hoomd...@googlegroups.com

Teresa,

Are you able to get any set of parameters (k or r0) working for the bonds? Can you attach the output of your script for a little more information?

-Ryan

Teresa Ruiz Herrero

unread,

Sep 29, 2014, 4:11:25 PM9/29/14

to hoomd...@googlegroups.com, rma...@umich.edu

I tried but I did not find any set of parameters that work for the fene.

This is the error file I got, it is just the system blowing up:

File Edit Options Buffers Tools Help
*Warning*: Trying to reset an uninitialized system
**ERROR**: Particle 730 is no longer in the simulation box.

**ERROR**: Cartesian coordinates:
**ERROR**: x: 77.3832 y: 27.9379 z: 4.5263
**ERROR**: Fractional coordinates:
**ERROR**: f.x: 1.48422 f.y: 0.855337 f.z: 0.556579
**ERROR**: Local box lo: (-39.3119, -39.3119, -40)
**ERROR**:           hi: (39.3119, 39.3119, 40)
Traceback (most recent call last):
File "mprnew.hoomd", line 124, in <module>
    run(runlen)
File "/share/apps/scisoft/GNU-4.4.7-compilations/HOOMD/1.0.1_CUDA/bin/../lib/hoomd/python-module/hoomd_script/__init__.py", line 242, in run
    globals.system.run(int(tsteps), callback_period, callback, limit_hours, int(limit_multiple));
RuntimeError: std::exception


and this is the output file:

mprnew.hoomd:019 | system = init.read_xml(filename="init0.xml")
HOOMD-blue is running on the following GPU(s):
[0]           Tesla M2050 14 SM_2.0 @ 1.15 GHz, 3071 MiB DRAM
notice(2): Reading init0.xml...
notice(2): --- hoomd_xml file read summary
notice(2): 29400 positions at timestep 0
notice(2): 29400 velocities
notice(2): 4 particle types
notice(2): 29400 bonds
notice(2): Group "all" created containing 29400 particles
mprnew.hoomd:029 | wca = pair.lj(cut1)
mprnew.hoomd:031 | wca.set_params(mode="shift")
mprnew.hoomd:034 | wca.pair_coeff.set(['MA','RA'], ['MA','MB','RA','RB'], epsilon=1.0, sigma=0.95,r_cut=0.95*atom_cut)
mprnew.hoomd:035 | wca.pair_coeff.set(['MB','RB'],['MB','RB'], epsilon=1.0, sigma=1.0,r_cut=atom_cut)
mprnew.hoomd:055 | attTab = pair.table(width = 100000,name='hydro')
mprnew.hoomd:058 | attTab.pair_coeff.set('MB','MB', func = hydrophobic, rmin = 0.0,rmax=(2**(1.0/6.0)+1.5),coeff = hyp_coeff)
mprnew.hoomd:059 | attTab.pair_coeff.set('RB','RB', func = hydrophobic, rmin = 0.0,rmax=(2**(1.0/6.0)+1.5),coeff = hyp_coeff)
mprnew.hoomd:060 | attTab.pair_coeff.set('RB','MB', func = hydrophobic, rmin = 0.0,rmax=(2**(1.0/6.0)+1.5),coeff = hyp_coeff_b)
mprnew.hoomd:062 | attTab.pair_coeff.set(['MA','RA'],['MA','MB','RA','RB'],func = hydrophobic,rmin = 0.0,rmax = 0.01,coeff = hypzero_coeff)
mprnew.hoomd:070 | fene = bond.fene()
mprnew.hoomd:071 | fene.bond_coeff.set('fene1', k=30.0, r0=1.5, sigma=0.95, epsilon= 1.0)    #fene.bond_coeff.set('fene1', k=30.0, r0=1.5, sigma=0.95, epsilon= 1.0)
mprnew.hoomd:072 | fene.bond_coeff.set('fene2', k=30.0, r0=1.5, sigma=1.0, epsilon= 1.0)    #fene.bond_coeff.set('fene2', k=0.5, r0=1.5, sigma=1.0, epsilon= 1.0)
mprnew.hoomd:073 | fene.bond_coeff.set('longbond', k=0, r0=0.0, sigma=0.0, epsilon=0)
mprnew.hoomd:077 | harmonic = bond.harmonic()
mprnew.hoomd:078 | harmonic.bond_coeff.set('longbond', k=10.0, r0=4.0)   #harmonic.bond_coeff.set('longbond', k=10.0, r0=4.0)
mprnew.hoomd:079 | harmonic.bond_coeff.set('fene1', k=0, r0=1.0)
mprnew.hoomd:080 | harmonic.bond_coeff.set('fene2', k=0, r0=1.0)
mprnew.hoomd:085 | all = group.all()
mprnew.hoomd:090 | nlist.reset_exclusions(exclusions=['bond'])
notice(2): -- Neighborlist exclusion statistics -- :
notice(2): Particles with 2 exclusions             : 29400
notice(2): Neighbors excluded by diameter (slj)    : no
notice(2): Neighbors excluded when in the same body: no
mprnew.hoomd:099 | imode = integrate.mode_standard(dt=ts)
mprnew.hoomd:101 | nvt = integrate.nvt(group=all, T=1.1,tau=0.4)
mprnew.hoomd:103 | analyze.log(filename=stringen, quantities=['potential_energy', 'kinetic_energy','temperature','pressure','volume'],
                        period=1000, header_prefix='#')
mprnew.hoomd:106 | analyze.log(filename=stringen2, quantities=['pressure','pressure_xx','pressure_yy','pressure_zz'],
                        period=1000, header_prefix='#')
mprnew.hoomd:109 | analyze.log(filename=stringpot, quantities=[ 'pair_lj_energy','pair_table_energy_hydro'],
                        period=1000, header_prefix='#')
mprnew.hoomd:114 | mol2 = dump.mol2()
mprnew.hoomd:115 | mol2.write(filename=stringmol2)
mprnew.hoomd:116 | dcd = dump.dcd(filename=stringdcd, period = anim_period)
mprnew.hoomd:119 | xml = dump.xml(filename=stringxml, period=1000000)
mprnew.hoomd:120 | xml.set_params(position=True, image=True, velocity=True, diameter=True, type=True, body=True, wall=True, bond=True, charge=True)
mprnew.hoomd:124 | run(runlen)
** starting run **

Jens Glaser

unread,

Oct 3, 2014, 10:29:38 AM10/3/14

to hoomd...@googlegroups.com

Tersa,

I think we will need a minimal reproducer script to understand the behavior better …

best - Jens

Teresa Ruiz

unread,

Oct 3, 2014, 10:50:21 AM10/3/14

to hoomd...@googlegroups.com

Here is my hoomd script and my initial configuration, is this enough?

Thank you very much!

Teresa

...

init0.xml

mpr.hoomd

Jens Glaser

unread,

Oct 26, 2014, 9:28:14 PM10/26/14

to hoomd...@googlegroups.com

Hi Teresa,

sorry in the long delay in getting back to you - the usual excuses for a busy schedule apply.

I noticed you have two bond types in your simulation, longbond and fene1/2.

fene = bond.fene()

fene.set_coeff('fene1', k=30.0, r0=1.5, sigma=0.95, epsilon= 1.0)    #fene.bond_coeff.set('fene1', k=30.0, r0=1.5, sigma=0.95, epsilon= 1.0)

fene.set_coeff('fene2', k=30.0, r0=1.5, sigma=1.0, epsilon= 1.0)

fene.set_coeff('longbond', k=0, r0=1.5, sigma=1.0, epsilon=0)

The spring constant k for longbond in the FENE potential is zero, but the maximum extension is finite. This seems to be

triggering the problem. Try setting r0 for that bond type to some large value (e.g. r0=10.5), to avoid a division by zero

in the force computation.

I’ll see if we can do a check against zero in the force evaluator for suppressing the rather

unhelpful error message you’re seeing.

Jens

--
You received this message because you are subscribed to the Google Groups "hoomd-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hoomd-users...@googlegroups.com.
To post to this group, send email to hoomd...@googlegroups.com.
Visit this group at http://groups.google.com/group/hoomd-users.
For more options, visit https://groups.google.com/d/optout.

<init0.xml><mpr.hoomd>

Teresa Ruiz

unread,

Nov 10, 2014, 2:55:50 PM11/10/14

to hoomd...@googlegroups.com

Hi Jens, thanks so much for getting back with the solution to the problem. I had already found another way to calculate things with the older version, but being able to use this version makes things way easier.

I tried what you said and it seems to work fine, I will make some runs to check that everything is ok.

Thank you again,

Teresa

Jens Glaser

unread,

Nov 10, 2014, 3:48:14 PM11/10/14

to hoomd...@googlegroups.com

BTW, the next 1.1 release will contain a fix to the problem of zero bond stiffness for FENE.

J

Reply all

Reply to author

Forward