mode2.out 18.8 KB
Newer Older
Jan Janssen's avatar
update  
Jan Janssen committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
 -------------------------------------------------------------
 ---------------------- Welcome to the -----------------------
     RuNNer Neural Network Energy Representation - RuNNer     
 ----------  (c) 2008-2020 Prof. Dr. Joerg Behler   ----------
 ----------  Georg-August-Universitaet Goettingen   ----------
 ----------           Theoretische Chemie           ----------
 ----------              Tammannstr. 6              ----------
 ----------        37077 Goettingen, Germany        ----------
 -------------------------------------------------------------
 -------------------------------------------------------------
  This program is free software: you can redistribute it and/or modify it
  under the terms of the GNU General Public License as published by the 
  Free Software Foundation, either version 3 of the License, or 
  (at your option) any later version.
    
  This program is distributed in the hope that it will be useful, but 
  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
  or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 
  for more details.
      
  You should have received a copy of the GNU General Public License along
  with this program. If not, see http://www.gnu.org/licenses. 
 -------------------------------------------------------------
 -------------------------------------------------------------
 When using RuNNer, please cite the following papers:
 J. Behler, Angew. Chem. Int. Ed. 56, 12828 (2017).
 J. Behler, Int. J. Quant. Chem. 115, 1032 (2015).
 -------------------------------------------------------------
 Whenever using high-dimensional NNPs irrespective of the Code please cite:
 J. Behler and M. Parrinello, Phys. Rev. Lett. 98, 146401 (2007).
 -------------------------------------------------------------
 The reference for the atom-centered symmetry functions is:
 J. Behler, J. Chem. Phys. 134, 074106 (2011).
 -------------------------------------------------------------
 For high-dimensional NNs including electrostatics:
 N. Artrith, T. Morawietz, and J. Behler, Phys. Rev. B 83, 153101 (2011).
 -------------------------------------------------------------
 -------------------------------------------------------------
 RuNNer has been written by Joerg Behler
  
 *** with contributions from some friends ***
  
 Tobias Morawietz - Nguyen Widrow weights and electrostatic screening
 Jovan Jose Kochumannil Varghese - Pair symmetry functions and pair NNPs
 Michael Gastegger and Philipp Marquetand - element decoupled Kalman filter
 Andreas Singraber - more efficient symmetry function type 3 implementation
 Sinja Klees and Mariana Rossi - some infrastructure for i-Pi compatibility
 Emir Kocer - Symmetry function groups
 Fenfei Wei and Emir Kocer - Hessian, frequencies and normal modes
 -------------------------------------------------------------
 -------------------------------------------------------------
 General job information:
 -------------------------------------------------------------
 Executing host    : tc-pc16.uni-pc.gwdg.
 User name         : mherbold            
 Starting date     : 28. 2.2021
 Starting time     : 15 h 19 min 
 Working directory : /scratch2/tmp/mherbold-onPC16SSD/df4/2                      
 -------------------------------------------------------------
 -------------------------------------------------------------
 Serial run requested
 -------------------------------------------------------------
 Reading control parameters from input.nn
 =============================================================
 -------------------------------------------------------------
 =============================================================
 General input parameters:
 -------------------------------------------------------------
 Short range NN is on
 Electrostatic NN is off
 vdW corrections switched off
 -------------------------------------------------------------
 RuNNer nn_type_short                                   1
 RuNNer is started in mode for fitting (2)
 debugging mode is                                        F
 parallelization mode                                     1
 enable detailed time measurement                         F
 using symmetry function groups                           F
 enable detailed time measurement at epoch level          F
 silent mode                                              F
 NN force check                                           F
 number of elements                                       1
 elements (sorted):
 29 Cu
 seed for random number generator                        20
 random number generator type                             6
 remove free atom reference energies                      F
 shortest allowed bond in structure                   0.500
 Cutoff_type for symmetry function is                     1
 Cutoff_alpha for inner cutoff radius is              0.000
 -------------------------------------------------------------
 Short range NN specifications:
 -------------------------------------------------------------
 global hidden layers short range NN                      2
 global nodes hidden layers short NN                15   15
 global activation functions short                      ttl
 -------------------------------------------------------------
 General fitting parameters:
 -------------------------------------------------------------
 number of fitting epochs                                10
 print date and time for each epoch                       F
 number of data sets in memory                          200
 Fitting mode 1 (online learning) selected         
 Randomly mixing all points in training set               T
 save Kalman filter data                                  F
 restart from old Kalman filter data                      F
 rescale symmetry functions                               T
 min value of scaled short range symmetry functions      0.000
 max value of scaled short range symmetry functions      1.000
 remove CMS from symmetry functions                       T
 calculate symmetry function correlation                  F
 weight analysis                                          F
 environment analysis                                     F
 find contradictions                                      F
 fix some weights                                         F
 using growth mode for fitting                            F
 global fit of short and charge NN (not implemented)      F
 error unit for fitting                                  eV
 Reading formatted files 
 Writing formatted files 
 Resetting Kalman filter matrices each epoch              F
 Preconditioning of weights is switched on
 -------------------------------------------------------------
 Fitting parameters short range part:
 -------------------------------------------------------------
 using forces for fitting                                 T
 using Kalman filter optimization (1) for short range energy
 using Kalman filter optimization (1) for short range forces
 short energy error threshold                    0.10000000
 short force error threshold                     1.00000000
 Kalman lambda (short)                           0.98000000
 Kalman nue (short)                              0.99870000
 use_noisematrix                                          F
 Kalman damp (short energy)                      1.00000000
 Kalman damp (short force)                       1.00000000
 restart fit with old weights (short)                     F
 automatic scaling factor for force update selected
 grouping energies in blocks of                           1
 fraction of energies used for update                 1.000
 grouping forces in blocks of                             1
 fraction of forces used for update                   0.010
 weights_min                                         -1.000
 weights_max                                          1.000
 Using repeated energy updates after each force update
 max_energy                                       10000.000
 max force component used for fitting             10000.000 Ha/Bohr
 noise energy threshold                          0.00000000 Ha/atom
 noise force threshold                           0.00000000 Ha/Bohr
 restart fit with old weights (charge)                    F
 -------------------------------------------------------------
 Fitting output options:
 -------------------------------------------------------------
 write weights in every epoch                             1
 write temporary weights each epoch                       F
 write trainpoints.out and testpoints.out                 T
 write binding energies only                              F
 write trainforces.out and testforces.out                 T
 =============================================================
 Element pairs:   1    , shortest distance (Bohr)
 pair    1 Cu Cu      2.824
 =============================================================
 => short range NN weights type 1                 Cu       466
 -------------------------------------------------------------
 -------------------------------------------------
 Atomic short range NN for element: Cu
 architecture       13   15   15    1
 -------------------------------------------------
   1   G  t  t  l
   2   G  t  t   
   3   G  t  t   
   4   G  t  t   
   5   G  t  t   
   6   G  t  t   
   7   G  t  t   
   8   G  t  t   
   9   G  t  t   
  10   G  t  t   
  11   G  t  t   
  12   G  t  t   
  13   G  t  t   
  14      t  t   
  15      t  t   
 -------------------------------------------------------------
 -------------------------------------------------------------
  short range atomic symmetry functions element Cu :
 -------------------------------------------------------------
    1 Cu  2  Cu              0.000   0.000  12.000
    2 Cu  2  Cu              0.006   0.000  12.000
    3 Cu  2  Cu              0.016   0.000  12.000
    4 Cu  2  Cu              0.040   0.000  12.000
    5 Cu  2  Cu              0.109   0.000  12.000
    6 Cu  3  Cu Cu   0.000  -1.000   1.000  12.000
    7 Cu  3  Cu Cu   0.000   1.000   1.000  12.000
    8 Cu  3  Cu Cu   0.000  -1.000   2.000  12.000
    9 Cu  3  Cu Cu   0.000   1.000   2.000  12.000
   10 Cu  3  Cu Cu   0.000  -1.000   4.000  12.000
   11 Cu  3  Cu Cu   0.000   1.000   4.000  12.000
   12 Cu  3  Cu Cu   0.000  -1.000  16.000  12.000
   13 Cu  3  Cu Cu   0.000   1.000  16.000  12.000
 -------------------------------------------------------------
 =============================================================
 Short range symmetry function values for element Cu
 Training set:  min           max       average         range        stddev      range/stddev
   1     6.72393532   25.25447208   14.00001115   18.53053676    4.02011080    4.60945921
   2     5.03976308   19.37796546   10.59714429   14.33820238    3.11321693    4.60559053
   3     3.25577537   13.11073890    6.98744718    9.85496353    2.14525332    4.59384607
   4     1.25883453    6.21227127    3.05195579    4.95343674    1.07630994    4.60224009
   5     0.06568766    1.53442885    0.48718005    1.46874120    0.26713619    5.49809898
   6     2.42989188   42.06637435   12.30898440   39.63648248    7.96564228    4.97593051
   7     6.59284339  109.34982696   33.28285935  102.75698358   21.03610544    4.88479124
   8     0.78142670   19.11390444    5.22756080   18.33247774    3.61016236    5.07802030
   9     5.20079617   86.39735705   26.20143575   81.19656088   16.68008353    4.86787495
  10     0.15497639    7.72411752    1.91198556    7.56914113    1.46079520    5.18152107
  11     3.59136462   60.38094318   17.95854981   56.78957856   11.72399989    4.84387403
  12     0.00059457    1.66933490    0.34750355    1.66874033    0.32743124    5.09646040
  13     0.52030958   18.24849080    4.70826264   17.72818122    3.59567219    4.93042198
 -------------------------------------------------------------
 Energies in training set (Ha/atom):
                   Emin          Emax          average        stddev          range
 Eshort         -0.135967      -0.091874      -0.118413       0.012680       0.044093
 Eelec           0.000000       0.000000       0.000000       0.000000       0.000000
 Etot           -0.135967      -0.091874      -0.118413       0.012680       0.044093
 -------------------------------------------------------------
 Energies in training set (eV/atom):
                   Emin          Emax          average        stddev          range
 Eshort         -3.699790      -2.499974      -3.222125       0.345036       1.199816
 Eelec           0.000000       0.000000       0.000000       0.000000       0.000000
 Etot           -3.699790      -2.499974      -3.222125       0.345036       1.199816
 -------------------------------------------------------------
 Force vectors in training set (Ha/Bohr):
               Fmin          Fmax          average        stddev          range
 Cu         0.000000       0.873705       0.011256       0.030995       0.873705
 -------------------------------------------------------------
 Force vectors in training set (eV/Bohr):
               Fmin          Fmax          average        stddev          range
 Cu         0.000000      23.774389       0.306275       0.843406      23.774389
 -------------------------------------------------------------
 Force components in training set (Ha/Bohr):
               Fmin          Fmax          range
 Cu         0.000000       0.873705       0.873705
 -------------------------------------------------------------
 Force components in training set (eV/Bohr):
               Fmin          Fmax          range
 Cu         0.000000      23.774389      23.774389
 -------------------------------------------------------------
 number of training points         7259
 number of training atoms         64729
 number of training forces       194187
 number of testing points           814
 number of testing atoms           7250
 number of testing forces         21750
 -------------------------------------------------------------
 Number of atoms for each element:   
            training:    testing:   
  1  Cu        64729        7250
 =============================================================
 Weight Preconditioner:
 Warning: Forces are not used for preconditioning
 ----------------------
 -------------------------------------------------------------
 Final preconditioning of the output values:
 --------------------------------------------
 Minimum NN Eshort         -3.346222 Ha/atom
 Minimum Ref Eshort        -0.135967 Ha/atom
 Maximum NN Eshort         -2.012861 Ha/atom
 Maximum Ref Eshort        -0.091874 Ha/atom
 Average NN Eshort         -2.855668 Ha/atom
 Average Ref Eshort        -0.118413 Ha/atom
 Stddev NN Eshort           0.200995 Ha/atom
 Stddev Ref Eshort          0.012680 Ha/atom
 Factor for connecting short range weights:       0.063086
 =============================================================
 -------------------------------------------------------------
 initialization time (min):    0.05
 -------------------------------------------------------------
 Did you check your output file for warnings? ;-)             
 -------------------------------------------------------------
 Short range energies below      10000.000 Ha/atom are used for fitting and Eshort RMSE!
 => Fitted energy range has width of          0.044 Ha/atom =          1.200 eV/atom
 => Number of short range training energies below max_energy:       7259
 Force components below      10000.000 Ha/Bohr are used for fitting and Fshort RMSE!
 Cu  => Fitted force range has width of          0.874 Ha/Bohr =         23.774 eV/Bohr
 Cu  => Number of short range training force components below max_force:     194187
 -------------------------------------------------------------------------------
 RMSEs (energies: eV/atom, forces: eV/Bohr):
                      --- E_short: ---    - time -
                          /atom              min
        epoch         train         test
 ENERGY     0      0.486020     0.481254    9.86
 FORCES     0      0.543702     0.502894
 -------------------------------------------------------------------------------
 ENERGY     1      0.039459     0.039840   19.05
 FORCES     1      0.201312     0.174885
 INFORMATION USED FOR UPDATE (E,F)     1      1998        45
 -------------------------------------------------------------------------------
 ENERGY     2      0.024635     0.026306   19.14
 FORCES     2      0.132738     0.123616
 INFORMATION USED FOR UPDATE (E,F)     2      5565       112
 -------------------------------------------------------------------------------
 ENERGY     3      0.022316     0.024581   19.13
 FORCES     3      0.120274     0.111427
 INFORMATION USED FOR UPDATE (E,F)     3      6033       131
 -------------------------------------------------------------------------------
 ENERGY     4      0.021333     0.023145   19.16
 FORCES     4      0.113496     0.105447
 INFORMATION USED FOR UPDATE (E,F)     4      6132       142
 -------------------------------------------------------------------------------
 ENERGY     5      0.022327     0.023597   19.13
 FORCES     5      0.113152     0.102596
 INFORMATION USED FOR UPDATE (E,F)     5      6064       137
 -------------------------------------------------------------------------------
 ENERGY     6      0.021007     0.022555   19.15
 FORCES     6      0.102685     0.094464
 INFORMATION USED FOR UPDATE (E,F)     6      6094       168
 -------------------------------------------------------------------------------
 ENERGY     7      0.021018     0.022213   19.15
 FORCES     7      0.098023     0.097181
 INFORMATION USED FOR UPDATE (E,F)     7      6226       158
 -------------------------------------------------------------------------------
 ENERGY     8      0.020692     0.022248   19.15
 FORCES     8      0.095995     0.097202
 INFORMATION USED FOR UPDATE (E,F)     8      6186       183
 -------------------------------------------------------------------------------
 ENERGY     9      0.020880     0.022219   19.16
 FORCES     9      0.094960     0.095833
 INFORMATION USED FOR UPDATE (E,F)     9      6122       176
 -------------------------------------------------------------------------------
 ENERGY    10      0.021217     0.022457   19.41
 FORCES    10      0.097554     0.094895
 INFORMATION USED FOR UPDATE (E,F)    10      6226       203
 =============================================================
 Best short range fit has been obtained in epoch     7
                    --- E_short: ---          --- F_short: ---
                   train         test        train         test
 OPTSHORT        0.021018     0.022213     0.098023     0.097181
 -------------------------------------------------------------
 max Eshort error in last epoch (train set):       0.281291  eV/atom (structure      788 )
 max Eshort error in last epoch (test set) :       0.261851  eV/atom (structure      253 )
 -------------------------------------------------------------
 Total runtime (s)  :      12095.013
 Total runtime (min):        201.584
 Total runtime (h)  :          3.360
 Normal termination of RuNNer
 -------------------------------------------------------------