Zero pressures in LAMMPS minimization #399

MusannaGalib · 2024-01-05T06:50:11Z

MusannaGalib
Jan 5, 2024

Hello Nequip Developers,

I trained a model from DFT data with extxyz dataset with StressForceOutput. However, during minimization, all pressure components are zero, and cell parameters and volumes are not changing. Please give your suggestions on this.

LAMMPS (29 Sep 2021 - Update 2)
OMP_NUM_THREADS environment is not set. Defaulting to 1 thread. (src/comm.cpp:98)
  using 1 OpenMP thread(s) per MPI task
Reading data file ...
  triclinic box = (0.0000000 0.0000000 0.0000000) to (24.024723 24.708023 21.072488) with tilt (0.0000000 0.0000000 0.0000000)
  1 by 1 by 1 MPI processor grid
  reading atoms ...
  1500 atoms
  read_data CPU = 0.032 seconds
NEQUIP is using device cpu
NequIP Coeff: type 1 is element Al
NequIP Coeff: type 2 is element O
Loading model from best_model.pth
Freezing TorchScript model...
Neighbor list info ...
  update every 1 steps, delay 0 steps, check yes
  max neighbors/atom: 2000, page size: 100000
  master list distance cutoff = 5
  ghost atom cutoff = 5
  1 neighbor lists, perpetual/occasional/extra = 1 0 0
  (1) pair nequip, perpetual
      attributes: full, newton off
      pair build: full/nsq
      stencil: none
      bin: none
Setting up cg style minimization ...
  Unit style    : metal
  Current step  : 0
WARNING: Energy due to 3 extra global DOFs will be included in minimizer energies
Per MPI rank memory allocation (min/avg/max) = 4.270 | 4.270 | 4.270 Mbytes
Step Temp PotEng Press Pxx Pyy Pzz Pxy Pxz Pyz Lx Ly Lz Volume 
       0            0   -10305.033            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
       1            0   -10306.602            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
       2            0   -10308.177            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
       3            0   -10310.239            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
       4            0    -10312.34            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
       5            0   -10314.786            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
       6            0   -10317.527            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
       7            0   -10320.302            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
       8            0   -10323.106            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
       9            0     -10326.2            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      10            0   -10329.262            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      11            0   -10332.177            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      12            0   -10335.042            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      13            0   -10337.573            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      14            0   -10340.254            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      15            0   -10342.744            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      16            0   -10345.254            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      17            0   -10347.784            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      18            0   -10350.176            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      19            0    -10352.94            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      20            0   -10355.538            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      21            0   -10358.407            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      22            0   -10361.399            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      23            0   -10364.793            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      24            0   -10368.145            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      25            0   -10372.004            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      26            0   -10376.164            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      27            0   -10380.688            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      28            0    -10385.52            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      29            0   -10390.985            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      30            0   -10396.815            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      31            0   -10403.229            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      32            0   -10410.569            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      33            0    -10418.82            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      34            0   -10428.142            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      35            0   -10438.979            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      36            0   -10450.791            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      37            0   -10461.882            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      38            0   -10472.052            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      39            0   -10481.242            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      40            0   -10489.296            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      41            0   -10496.229            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      42            0   -10502.191            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      43            0   -10507.252            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      44            0   -10511.555            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      45            0   -10515.241            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      46            0   -10518.463            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      47            0   -10521.353            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      48            0   -10523.985            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      49            0   -10526.417            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      50            0   -10528.688            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      51            0   -10530.751            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      52            0   -10532.548            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      53            0   -10533.962            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      54            0   -10534.867            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      55            0   -10535.251            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      56            0   -10535.522            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      57            0   -10535.914            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      58            0   -10536.481            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      59            0   -10537.229            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      60            0   -10538.126            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      61            0   -10539.194            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      62            0   -10540.404            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      63            0   -10541.758            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      64            0   -10543.248            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      65            0   -10544.881            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      66            0   -10546.642            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      67            0   -10548.553            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      68            0   -10550.616            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      69            0   -10552.818            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      70            0   -10555.211            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      71            0   -10557.777            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      72            0   -10560.534            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      73            0   -10563.522            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      74            0   -10566.748            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      75            0    -10570.24            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      76            0   -10573.989            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      77            0   -10577.876            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      78            0   -10581.911            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      79            0   -10586.103            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      80            0   -10590.479            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      81            0   -10595.064            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      82            0   -10599.871            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      83            0   -10604.963            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      84            0   -10610.329            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      85            0   -10616.063            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      86            0   -10622.122            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      87            0   -10628.312            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      88            0   -10634.657            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      89            0   -10640.927            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      90            0   -10647.146            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      91            0   -10653.323            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      92            0   -10659.506            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      93            0   -10665.694            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      94            0   -10671.902            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      95            0   -10678.162            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      96            0    -10684.49            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      97            0   -10690.856            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      98            0   -10697.324            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
      99            0   -10703.861            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
     100            0   -10710.508            0            0            0            0            0            0            0    24.024723    24.708023    21.072488    12508.701 
Loop time of 499.153 on 1 procs for 100 steps with 1500 atoms

My data file looks like this -

20
virial="261.83204 0.0 0.0 0.0 271.97767 0.0 0.0 0.0 105.90791" Lattice="4.804944559452511 0.0 0.0 0.0 4.941604551830464 0.0 0.0 0.0 7.024162707093621" Properties=species:S:1:pos:R:3:forces:R:3 energy=-95.1057827 pbc="T T T"
Al       0.96983000       0.12412000       2.73934000       4.30521200       0.61509100      -3.60469100
Al       0.96983000       1.85252000       4.28482000       4.30521200      -0.61509100       3.60469100
Al       0.95215000       2.10076000       0.77274000      -4.30521200       0.61509100       3.60469100
Al       0.95215000       3.82916000       6.25142000      -4.30521200      -0.61509100      -3.60469100
Al       2.87412000       3.82916000       4.28482000      -4.30521200      -0.61509100       3.60469100
Al       2.87412000       2.10076000       2.73934000      -4.30521200       0.61509100      -3.60469100
Al       2.89181000       1.85252000       6.25142000       4.30521200      -0.61509100      -3.60469100
Al       2.89181000       0.12412000       0.77274000       4.30521200       0.61509100       3.60469100
O        0.41158000       0.40478000       4.59484000       5.09482300      -1.22603600       4.01148700
O        0.41158000       1.57187000       2.42932000       5.09482300       1.22603600      -4.01148700
O        1.51039000       2.38142000       5.94140000      -5.09482300      -1.22603600      -4.01148700
O        1.51039000       3.54851000       1.08276000      -5.09482300       1.22603600       4.01148700
O        3.43237000       3.54851000       2.42932000      -5.09482300       1.22603600      -4.01148700
O        3.43237000       2.38142000       4.59484000      -5.09482300      -1.22603600       4.01148700
O        2.33356000       1.57187000       1.08276000       5.09482300       1.22603600       4.01148700
O        2.33356000       0.40478000       5.94140000       5.09482300      -1.22603600      -4.01148700
O        0.19292000       0.98832000       0.00000000       2.18218700      -0.00000000       0.00000000
O        1.72906000       2.96496000       3.51208000      -2.18218700       0.00000000       0.00000000
O        3.65103000       2.96496000       0.00000000      -2.18218700       0.00000000       0.00000000
O        2.11490000       0.98832000       3.51208000       2.18218700      -0.00000000       0.00000000
18
virial="-50.43675 0.7527 0.69011 0.7527 -54.0229 3.10856 0.69011 3.10856 -31.9972" Lattice="4.8617224778965955 0.0 0.0 0.0 4.936388270474029 0.0 0.0 0.0 7.01702523145114" Properties=species:S:1:pos:R:3:forces:R:3 energy=-123.60290169 pbc="T T T"
Al       1.29070000       0.22316000       2.48354000      -0.74630000       0.06558700       1.07988300
Al       1.16988000       2.90339000       4.32268000      -1.13841800       0.78274300       0.33374100
Al       1.42322000       2.95428000       0.82270000       0.71506300      -0.82486100      -0.56865400
Al       1.35726000       5.61243000       6.31182000      -0.06242500       0.62717100      -0.13548400
Al       4.16819000       5.35547000       4.29953000      -0.10530200      -0.86839100       0.44546700
Al       4.09544000       2.67851000       6.28576000      -0.02284700       0.27119100       0.56704500
Al       4.17440000       0.07315000       0.81082000      -0.01752300       0.37632300      -0.82360800
O        0.55956000       0.72590000       4.71245000       0.32977500      -0.72975300       0.20597700
O        1.28526000       2.28191000       2.60818000       0.06834700      -0.98058700      -0.05937700
O        2.07737000       3.51566000       5.92790000      -0.62916300      -0.58965100       0.23823700
O        2.16029000       4.98235000       1.09465000      -0.59612300       0.92347100      -0.60103000
O        4.79833000       5.54073000       2.56538000       1.04645000       0.49345300       0.00156400
O        4.80327000       3.42246000       4.67881000       0.71953300       0.43828200       0.21955400
O        3.26435000       1.98524000       0.86648000       0.08046500      -0.29781100      -1.10741000
O        3.36685000       0.65375000       5.70696000       0.34615100      -0.84529900      -0.11154300
O        0.30955000       1.33441000       0.12921000       0.16932600      -0.17940100       0.52092500
O        2.23393000       4.76045000       3.69953000      -0.26801700       0.97025500      -0.45606100
O        5.19750000       4.17788000       0.12554000       0.11100700       0.36727700       0.25077500

training and validation loss plots-

I am also adding the full.yaml here-

# IMPORTANT: READ THIS

# This is a full yaml file with all nequip options.
# It is primarily intented to serve as documentation/reference for all options
# For a simpler yaml file containing all necessary features to get you started, we strongly recommend to start with configs/example.yaml

# Two folders will be used during the training: 'root'/process and 'root'/'run_name'
# run_name contains logfiles and saved models
# process contains processed data sets
# if 'root'/'run_name' exists, 'root'/'run_name'_'year'-'month'-'day'-'hour'-'min'-'s' will be used instead.
root: results/alumina
run_name: example-run-alumina
seed: 123                                                                         # model seed
dataset_seed: 456                                                                 # data set seed
append: true                                                                      # set true if a restarted run should append to the previous log file
default_dtype: float32                                                            # type of float to use, e.g. float32 and float64
allow_tf32: false                                                                 # whether to use TensorFloat32 if it is available
# device:  cuda                                                                   # which device to use. Default: automatically detected cuda or "cpu"

# == network ==

# `model_builders` defines a series of functions that will be called to construct the model
# each model builder has the opportunity to update the model, the config, or both
# model builders from other packages are allowed (see mir-group/allegro for an example); those from `nequip.model` don't require a prefix
# these are the default model builders:
model_builders:
 - SimpleIrrepsConfig         # update the config with all the irreps for the network if using the simplified `l_max` / `num_features` / `parity` syntax
 - EnergyModel                # build a full NequIP model
 - PerSpeciesRescale          # add per-atom / per-species scaling and shifting to the NequIP model before the total energy sum
 #- ForceOutput                # wrap the energy model in a module that uses autodifferention to compute the forces
 - StressForceOutput
 - RescaleEnergyEtc           # wrap the entire model in the appropriate global rescaling of the energy, forces, etc.
 - initialize_from_state 
initial_model_state: /home/galibubc/scratch/ML_LAMMPS/ML/results_full_model_withstress_trial_2/alumina/example-run-alumina/best_model.pth

#   ^ global rescaling blocks must always go last!

r_max: 4.0                                                                        # cutoff radius in length units, here Angstrom, this is an important hyperparamter to scan
num_layers: 4                                                                     # number of interaction blocks, we find 3-5 to work best

l_max: 1                                                                          # the maximum irrep order (rotation order) for the network's features, l=1 is a good default, l=2 is more accurate but slower
parity: true                                                                      # whether to include features with odd mirror parityy; often turning parity off gives equally good results but faster networks, so do consider this
num_features: 32                                                                  # the multiplicity of the features, 32 is a good default for accurate network, if you want to be more accurate, go larger, if you want to be faster, go lower

# alternatively, the irreps of the features in various parts of the network can be specified directly:
# the following options use e3nn irreps notation
# either these four options, or the above three options, should be provided--- they cannot be mixed.
# chemical_embedding_irreps_out: 32x0e                                              # irreps for the chemical embedding of species
# feature_irreps_hidden: 32x0o + 32x0e + 32x1o + 32x1e                              # irreps used for hidden features, here we go up to lmax=1, with even and odd parities; for more accurate but slower networks, use l=2 or higher, smaller number of features is faster
# irreps_edge_sh: 0e + 1o                                                           # irreps of the spherical harmonics used for edges. If a single integer, indicates the full SH up to L_max=that_integer
# conv_to_output_hidden_irreps_out: 16x0e                                           # irreps used in hidden layer of output block


nonlinearity_type: gate                                                           # may be 'gate' or 'norm', 'gate' is recommended
resnet: false                                                                     # set true to make interaction block a resnet-style update
                                                                                  # the resnet update will only be applied when the input and output irreps of the layer are the same

# scalar nonlinearities to use — available options are silu, ssp (shifted softplus), tanh, and abs.
# Different nonlinearities are specified for e (even) and o (odd) parity;
# note that only tanh and abs are correct for o (odd parity).
# silu typically works best for even 
nonlinearity_scalars:
  e: silu
  o: tanh

nonlinearity_gates:
  e: silu
  o: tanh

# radial network basis
num_basis: 8                                                                      # number of basis functions used in the radial basis, 8 usually works best
BesselBasis_trainable: true                                                       # set true to train the bessel weights
PolynomialCutoff_p: 6                                                             # p-exponent used in polynomial cutoff function, smaller p corresponds to stronger decay with distance

# radial network
invariant_layers: 2                                                               # number of radial layers, usually 1-3 works best, smaller is faster
invariant_neurons: 64                                                             # number of hidden neurons in radial function, smaller is faster
avg_num_neighbors: auto                                                           # number of neighbors to divide by, null => no normalization, auto computes it based on dataset 
use_sc: true                                                                      # use self-connection or not, usually gives big improvement

# to specify different parameters for each convolutional layer, try examples below
# layer1_use_sc: true                                                             # use "layer{i}_" prefix to specify parameters for only one of the layer,
# priority for different definitions:
#   invariant_neurons < InteractionBlock_invariant_neurons < layer{i}_invariant_neurons

# data set
# there are two options to specify a dataset, npz or ase
# npz works with npz files, ase can ready any format that ase.io.read can read
# in most cases working with the ase option and an extxyz file is by far the simplest way to do it and we strongly recommend using this
# simply provide a single extxyz file that contains the structures together with energies and forces (generated with ase.io.write(atoms, format='extxyz', append=True))



chemical_symbols:
  - Al
  - O 





# As an alternative option to npz, you can also pass data ase ASE Atoms-objects
# This can often be easier to work with, simply make sure the ASE Atoms object
# has a calculator for which atoms.get_potential_energy() and atoms.get_forces() are defined
dataset: ase
dataset_file_name: nequip_data.extxyz                                                       # need to be a format accepted by ase.io.read
ase_args:                                                                        # any arguments needed by ase.io.read
   format: extxyz
dataset_include_keys:
   - virial
# If you want to use a different dataset for validation, you can specify
# the same types of options using a `validation_` prefix:
# validation_dataset: ase
# validation_dataset_file_name: xxx.xyz                                            # need to be a format accepted by ase.io.read

# logging
wandb: true                                                                        # we recommend using wandb for logging
wandb_project: alumina                                                     # project name used in wandb
wandb_watch: false

# see https://docs.wandb.ai/ref/python/watch
# wandb_watch_kwargs:
#   log: all
#   log_freq: 1
#   log_graph: true

verbose: info                                                                      # the same as python logging, e.g. warning, info, debug, error. case insensitive
log_batch_freq: 1                                                                  # batch frequency, how often to print training errors withinin the same epoch
log_epoch_freq: 1                                                                  # epoch frequency, how often to print 
save_checkpoint_freq: -1                                                           # frequency to save the intermediate checkpoint. no saving of intermediate checkpoints when the value is not positive.
save_ema_checkpoint_freq: -1                                                       # frequency to save the intermediate ema checkpoint. no saving of intermediate checkpoints when the value is not positive.

# training
n_train: 1400                                                                       # number of training data
n_val: 281                                                                          # number of validation data
learning_rate: 0.005                                                               # learning rate, we found values between 0.01 and 0.005 to work best - this is often one of the most important hyperparameters to tune
batch_size: 5                                                                      # batch size, we found it important to keep this small for most applications including forces (1-5); for energy-only training, higher batch sizes work better
validation_batch_size: 10                                                          # batch size for evaluating the model during validation. This does not affect the training results, but using the highest value possible (<=n_val) without running out of memory will speed up your training.
max_epochs: 1000                                                                 # stop training after _ number of epochs, we set a very large number here, it won't take this long in practice and we will use early stopping instead
train_val_split: random                                                            # can be random or sequential. if sequential, first n_train elements are training, next n_val are val, else random, usually random is the right choice
shuffle: true                                                                      # If true, the data loader will shuffle the data, usually a good idea
metrics_key: validation_loss                                                       # metrics used for scheduling and saving best model. Options: `set`_`quantity`, set can be either "train" or "validation, "quantity" can be loss or anything that appears in the validation batch step header, such as f_mae, f_rmse, e_mae, e_rmse
use_ema: true                                                                      # if true, use exponential moving average on weights for val/test, usually helps a lot with training, in particular for energy errors
ema_decay: 0.99                                                                    # ema weight, typically set to 0.99 or 0.999
ema_use_num_updates: true                                                          # whether to use number of updates when computing averages
report_init_validation: true                                                       # if True, report the validation error for just initialized model

# early stopping based on metrics values. 
# LR, wall and any keys printed in the log file can be used. 
# The key can start with Training or validation. If not defined, the validation value will be used.
early_stopping_patiences:                                                          # stop early if a metric value stopped decreasing for n epochs
  validation_loss: 50

early_stopping_delta:                                                              # If delta is defined, a decrease smaller than delta will not be considered as a decrease
  validation_loss: 0.005

early_stopping_cumulative_delta: false                                             # If True, the minimum value recorded will not be updated when the decrease is smaller than delta

early_stopping_lower_bounds:                                                       # stop early if a metric value is lower than the bound
  LR: 1.0e-5

early_stopping_upper_bounds:                                                       # stop early if a metric value is higher than the bound
  cumulative_wall: 1.0e+100

# loss function
loss_coeffs:                                                                       # different weights to use in a weighted loss functions
  forces: 1   
  virial: 1
  # if using PerAtomMSELoss, a default weight of 1:1 on each should work well
  total_energy:                                                                    
    - 1
    - PerAtomMSELoss
# note that the ratio between force and energy loss matters for the training process. One may consider using 1:1 with the PerAtomMSELoss. If the energy loss still significantly dominate the loss function at the initial epochs, tune the energy loss weight lower helps the training a lot.


# # default loss function is MSELoss, the name has to be exactly the same as those in torch.nn.
# the only supprted targets are forces and total_energy



# output metrics
metrics_components:
  - - forces                               # key 
    - mae                                  # "rmse" or "mae"
  - - forces
    - rmse
  - - forces
    - mae
    - PerSpecies: True                     # if true, per species contribution is counted separately
      report_per_component: False          # if true, statistics on each component (i.e. fx, fy, fz) will be counted separately
  - - forces                                
    - rmse                                  
    - PerSpecies: True                     
      report_per_component: False    
  - - total_energy
    - mae    
  - - total_energy
    - mae
    - PerAtom: True                        # if true, energy is normalized by the number of atoms

# optimizer, may be any optimizer defined in torch.optim
# the name `optimizer_name`is case sensitive
# IMPORTANT: for NequIP (not for Allegro), we find that in most cases AMSGrad strongly improves
# out-of-distribution generalization over Adam. We highly recommed trying both AMSGrad (by setting
# optimizer_amsgrad: true) and Adam (by setting optimizer_amsgrad: false)
optimizer_name: Adam                                                          
optimizer_amsgrad: true
optimizer_betas: !!python/tuple
  - 0.9
  - 0.999
optimizer_eps: 1.0e-08
optimizer_weight_decay: 0

# gradient clipping using torch.nn.utils.clip_grad_norm_
# see https://pytorch.org/docs/stable/generated/torch.nn.utils.clip_grad_norm_.html#torch.nn.utils.clip_grad_norm_
# setting to inf or null disables it
max_gradient_norm: null

# lr scheduler, currently only supports the two options listed below, if you need more please file an issue
# first: on-plateau, reduce lr by factory of lr_scheduler_factor if metrics_key hasn't improved for lr_scheduler_patience epoch
lr_scheduler_name: ReduceLROnPlateau
lr_scheduler_patience: 100
lr_scheduler_factor: 0.5


per_species_rescale_scales_trainable: false
# whether the scales are trainable. Defaults to False. Optional
per_species_rescale_shifts_trainable: false
# whether the shifts are trainable. Defaults to False. Optional
per_species_rescale_shifts: dataset_per_atom_total_energy_mean
# initial atomic energy shift for each species. default to the mean of per atom energy. Optional
# the value can be a constant float value, an array for each species, or a string
# string option include: 
# *  "dataset_per_atom_total_energy_mean", which computes the per atom average
# *  "dataset_per_species_total_energy_mean", which automatically compute the per atom energy mean using a GP model
per_species_rescale_scales: dataset_forces_rms
# initial atomic energy scale for each species. Optional.
# the value can be a constant float value, an array for each species, or a string
# string option include: 
# *  "dataset_per_atom_total_energy_std", which computes the per atom energy std
# *  "dataset_per_species_total_energy_std", which uses the GP model uncertainty
# *  "dataset_per_species_forces_rms", which compute the force rms for each species
# If not provided, defaults to dataset_per_species_force_rms or dataset_per_atom_total_energy_std, depending on whether forces are being trained.
# per_species_rescale_kwargs: 
#   total_energy: 
#     alpha: 0.001
#     max_iteration: 20
#     stride: 100
# keywords for ridge regression decomposition of per specie energy. Optional. Defaults to 0.001. The value should be in the range of 1e-3 to 1e-2
# per_species_rescale_arguments_in_dataset_units: True
# if explicit numbers are given for the shifts/scales, this parameter must specify whether the given numbers are unitless shifts/scales or are in the units of the dataset. If ``True``, any global rescalings will correctly be applied to the per-species values.

# global energy shift and scale
# When "dataset_total_energy_mean", the mean energy of the dataset. When None, disables the global shift. When a number, used directly.
# Warning: if this value is not None, the model is no longer size extensive
global_rescale_shift: null

# global energy scale. When "dataset_force_rms", the RMS of force components in the dataset. When "dataset_total_energy_std", the stdev of energies in the dataset. When null, disables the global scale. When a number, used directly.
# If not provided, defaults to either dataset_force_rms or dataset_total_energy_std, depending on whether forces are being trained.
global_rescale_scale: dataset_forces_rms

# whether the shift of the final global energy rescaling should be trainable
global_rescale_shift_trainable: false

# whether the scale of the final global energy rescaling should be trainable
global_rescale_scale_trainable: false

MusannaGalib · 2024-01-09T01:40:49Z

MusannaGalib
Jan 9, 2024
Author

@Linux-cpp-lisp @simonbatzner can you please comment on this?

3 replies

Linux-cpp-lisp Jan 9, 2024
Maintainer

What version of pair_nequip or pair_allegro are you using?

MusannaGalib Jan 9, 2024
Author

The latest one of pair_nequip - 0.5.2

MusannaGalib Jan 11, 2024
Author

@Linux-cpp-lisp, any update?

I tried with the fcu dataset it also shows zero pressures.

Also, using npt it shows non-physical temperatures.

Step Lx Ly Lz Press Pxx Pyy Pzz PotEng Temp 
    7420    54.682383    56.237634    79.938062     13.49638    20.457558    9.1481635    10.883418   -127.16474    1264.7597 
    7430    54.859559    56.419849    80.197069     13.29342    17.569208    6.9173794    15.393672   -127.23701    1257.8884 
    7440    55.037017    56.602354    80.456488    14.492117    15.309807    6.2115781    21.954966   -127.60444    1384.6658 
    7450    55.215291    56.785698    80.717099    16.397972    14.894423    5.8716547    28.427837   -128.22218    1582.0371 
    7460    55.395151    56.970674     80.98003    17.545658    14.508404    6.9964019    31.132167   -128.52472    1709.3593 
    7470    55.577141     57.15784    81.246074    18.534316    16.173737    9.8546924    29.574519   -128.79716     1823.533 
    7480    55.761708    57.347657    81.515885    19.167501    16.882058    13.160083    27.460361   -128.98928    1904.6805 
    7490    55.949053     57.54033    81.789758    17.509722    16.287701    14.820164    21.421301   -128.60815    1757.5429 
    7500    56.138535    57.735201    82.066755    15.460892    15.329239    16.342088    14.711347   -128.09485    1567.7118 
    7510    56.329538    57.931636    82.345973     18.99706    19.261188    21.519834    16.210157   -128.96899    1946.0025 
    7520    56.523452    58.131066     82.62945    22.648411    21.453239    27.227167    19.264826     -129.939    2344.0789 
    7530    56.721733    58.334986    82.919309    21.115239    24.757444    25.164808    13.423465   -129.68524    2208.4772 
    7540    56.923956     58.54296     83.21493     20.65527     24.00212    24.166492    13.797199   -129.76387    2183.5571 
    7550    57.129832    58.754692    83.515893    20.064788    22.693563    23.978246    13.522555   -129.52972    2144.2324 
    7560    57.339035    58.969845    83.821718    17.604458    14.105681    23.336912    15.370782   -128.92375    1902.0514 
    7570    57.550449    59.187272    84.130777    15.949153    6.3287099    19.743347    21.775401   -128.49843    1742.3372 
    7580    57.763576    59.406461    84.442339    19.994283    8.3155998    24.001154    27.666096   -129.62623    2208.5972 
    7590    57.979877    59.628914    84.758541    18.458301    11.099683    19.881161    24.394061   -129.34655    2061.9212 
    7600    58.198968    59.854236    85.078822    18.380012    14.883543    15.355799    24.900694   -129.59441     2076.539 
    7610    58.421146    60.082733    85.403615    25.189953    22.511372    23.399132    29.659355   -131.66457    2878.6312 
    7620    58.648973    60.317039    85.736666     23.21147    23.611701    16.101532    29.921178   -131.11447    2683.6899 
    7630    58.882474    60.557181    86.078012    28.501434    28.538528    18.958389    38.007385   -132.64232    3334.8269 
    7640    59.124299    60.805884    86.431527    53.385336    26.769507    35.214022    98.172479   -139.77594    6323.6589 
    7650     59.38671    61.075759    86.815135    79.873218    31.293681    75.953858    132.37211   -147.57862     9587.766 
    7660    59.681194    61.378619     87.24563     56.56841    59.691261    41.522214    68.491754   -141.31583    6891.8359 
    7670    60.003406    61.709995     87.71666     70.22523    66.974209     48.58894     95.11254    -146.6582     8694.995 
    7680    60.354836     62.07142    88.230402    63.194345    59.492097    49.588104    80.502832   -145.50706    7962.7465 
    7690    60.731895    62.459203     88.78161    84.365991     102.1918    73.726359    77.179817   -152.34945    10830.945 
    7700    61.143256    62.882264    89.382963    84.581945    127.44388     39.59583    86.706124   -152.53877    11080.818 
    7710    61.595496    63.347366    90.044074    97.959342    136.29591       55.837    101.74512   -157.79824    13120.222 
    7720    62.095462    63.861552    90.774955    89.367991    132.10361    73.585371    62.414991   -157.36355    12263.375 
    7730    62.642048    64.423684    91.573989    154.89584    175.84872    64.576603     224.2622   -182.68009    21821.573 
    7740    63.273088    65.072671    92.496481     204.4689    324.01171    134.58248    154.81252   -204.30015     29684.71 
    7750     64.01128    65.831859    93.575616    195.83622     166.2818    255.84109    165.38576   -206.05326    29438.183 
    7760    64.858097    66.702761    94.813545    329.83081    256.59751    535.17028    197.72464   -264.66681    51574.175 
    7770    65.857998    67.731101    96.275261    259.78389    197.90419    144.64325    436.80425   -243.48539    42529.105 
    7780    67.385674    69.302226    98.508511    16052.867    2177.1531    36302.829    9678.6181   -108.42469    2815164.5 
    7790    77.126653    79.320254    112.74847    6054.1619    809.81211    13707.153    3645.5205    -102.9187    1591900.6 
    7800     87.25388    89.735514    127.55307    2487.6134    338.71644    5624.6682    1499.4557   -107.88615    947077.66

I found this thread, which says a similar problem. But I checked, the modifications are already added in pair_quip.cpp inside ML-QUIP

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Zero pressures in LAMMPS minimization #399

{{title}}

Replies: 1 comment 3 replies

{{title}}

{{title}}

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

Select a reply

Zero pressures in LAMMPS minimization #399

MusannaGalib Jan 5, 2024

Replies: 1 comment · 3 replies

MusannaGalib Jan 9, 2024 Author

Linux-cpp-lisp Jan 9, 2024 Maintainer

MusannaGalib Jan 9, 2024 Author

MusannaGalib Jan 11, 2024 Author

MusannaGalib
Jan 5, 2024

Replies: 1 comment 3 replies

MusannaGalib
Jan 9, 2024
Author

Linux-cpp-lisp Jan 9, 2024
Maintainer

MusannaGalib Jan 9, 2024
Author

MusannaGalib Jan 11, 2024
Author