Project

General

Profile

Bug #3304

Failure in Log Output for Expanded Ensemble Weights using Simulated Tempering in GROMACS 2019.5

Added by Daniel Kozuch 9 months ago. Updated 7 months ago.

Status:
New
Priority:
Normal
Category:
Logging
Target version:
Affected version - extra info:
with GPU
Affected version:
Difficulty:
uncategorized
Close

Description

When using simulated tempering with metropolis transition the log output for the lambda weights is only 0, inf, or -nan (see log output example below). The mdp settings are shown below the log output. I can attach a .gro and .top file if need be, but I think this is not system dependent. I think there is also an error in the documentation, "lmc-mc-move" should be "lmc-move".

Log output:
--------------------------------------------

MC-lambda information
N Temp.(K) Count G(in kT) dG(in kT)
1 300.000 0 0.00000 0.00000
2 301.000 0 0.00000 0.00000
3 302.000 3 0.00000 inf
4 303.000 6 inf -nan
5 304.000 6 inf -nan <<
6 305.000 4 inf -nan
7 306.000 1 -nan -nan
8 307.000 0 -nan -nan
9 308.000 0 -nan -nan
10 309.000 0 -nan -nan
11 310.000 0 -nan 0.00000

MDP Settings:
----------------------------------------------------

integrator = md-vv
dt = 0.002
nstcomm = 100

nstlog = 500
nstcalcenergy = 100
nstenergy = 500
nstxtcout = 500
nstvout = 0
xtc-precision = 1000

cutoff-scheme = Verlet
nstlist = 10
ns-type = Grid
pbc = xyz
periodic-molecules = no
rlist = 1

coulombtype = PME
coulomb-modifier = Potential-shift-Verlet
rcoulomb = 1
vdw-type = Cut-off
vdw-modifier = Potential-shift-Verlet
rvdw = 1
DispCorr = EnerPres
optimize-fft = yes

tcoupl = v-rescale
nsttcouple = -1
nh-chain-length = 10
tc-grps = Protein Non-Protein
tau-t = 0.2 0.2
ref-t = 300 300
pcoupl = Berendsen
pcoupltype = Isotropic
nstpcouple = -1
tau-p = 2
compressibility = 0.000045
ref-p = 1 1 1 0 0 0

gen-vel = yes
gen-temp = 300
gen-seed = -1

constraints = h-bonds
constraint-algorithm = LINCS
continuation = no

; EXPANDED ENSEMBLE
simulated-tempering = yes
nstexpanded = 100
sim-temp-low = 300
sim-temp-high = 310
simulated-tempering-scaling = linear
lmc-stats = metropolis-transition
lmc-move = metropolis
temperature-lambdas = 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
init-lambda-state = 3
lmc-weights-equil = no
nstdhdl = 100

sim_anneal_for_redmine.tar.gz (1.06 MB) sim_anneal_for_redmine.tar.gz Daniel Kozuch, 02/28/2020 10:22 PM

History

#1 Updated by Paul Bauer 7 months ago

  • Target version changed from 2019.6 to 2020.2

sorry, but there are no inputs here for any of us to reproduce the bug
Please upload inputs and commandlines so we can have a chance to fix it. This is going to happen though in the 2020 branch, because 2019 is for critical fixes only now, or for issues that only affect the group scheme code that has been removed in 2020

#2 Updated by Daniel Kozuch 7 months ago

Paul Bauer wrote:

sorry, but there are no inputs here for any of us to reproduce the bug
Please upload inputs and commandlines so we can have a chance to fix it. This is going to happen though in the 2020 branch, because 2019 is for critical fixes only now, or for issues that only affect the group scheme code that has been removed in 2020

Here are a some files for reproducing in a zipped directory (attached):
sim.mdp
structure.gro
topol.top
sim.tpr

sim.tpr was build with "gmx grompp -f sim.mdp -c structure.gro -p topol.top -o sim.tpr"
system was run with "srun gmx mdrun -s sim -cpi sim -deffnm sim -ntomp 7" on a slurm node with 4 tasks, 7 cpus per task, and 4 gpus

#3 Updated by Erik Lindahl 7 months ago

Michael, will you look at this?

#4 Updated by Michael Shirts 7 months ago

Will put it on my todo list.

Also available in: Atom PDF