Gaussian SCF error in DP-GEN fp process

[paseri3739]

When running Gaussian16, an SCF error occurs due to incorrect structure. Since the error does not result in a normal completion, a completion flag file is not generated and the step cannot proceed any further.
Is it possible to ignore the structure that causes the SCF error? Or do I have to configure the SCF settings on the Gaussian16 side and use unreliable calculated values?

parameters:

type_map:

• Na
  init_data_prefix: ''
  init_data_sys:
• /home/sekinelabo/Documents/DeepMD-Sodium/dpgen/data/Na7/
  sys_format: deepmd/npy
  sys_configs:
• • /home/sekinelabo/Documents/DeepMD-Sodium/dpgen/data/Na7/
    numb_models: 4
    default_training_param:
    model:
    type_map:
  • Na
    data_stat_nbatch: 1
    type: standard
    descriptor:
    type: se_e2_a
    sel:
    • 60
      rcut: 6.0
      rcut_smth: 1.8
      neuron:
    • 25
    • 50
    • 100
      axis_neuron: 8
      resnet_dt: false
      precision: float64
      seed: 1
      fitting_net:
      type: ener
      neuron:
    • 120
    • 120
    • 120
      precision: float64
      resnet_dt: true
      seed: 1
      learning_rate:
      type: exp
      start_lr: 0.001
      stop_lr: 3.0e-08
      decay_steps: 5000
      loss:
      type: ener
      start_pref_e: 0.02
      limit_pref_e: 1
      start_pref_f: 1000
      limit_pref_f: 1
      start_pref_v: 0
      limit_pref_v: 0
      training:
      numb_steps: 1000000
      seed: 1
      disp_file: lcurve.out
      disp_freq: 100
      save_freq: 1000
      save_ckpt: model.ckpt
      disp_training: true
      time_training: true
      profiling: false
      profiling_file: timeline.json
      fp_task_max: 20
      fp_task_min: 1
      model_devi_engine: lammps
      model_devi_jobs:
• sys_idx:
  • 0
    temps:
  • 300
    press:
  • 1
    nsteps: 1000
    ensemble: nvt
    trj_freq: 10
    _idx: "00"
• sys_idx:
  • 0
    temps:
  • 300
    press:
  • 1
    nsteps: 3000
    ensemble: nvt
    trj_freq: 10
    _idx: "01"
    model_devi_dt: 0.002
    model_devi_skip: 0
    model_devi_f_trust_lo: 0.05
    model_devi_f_trust_hi: 0.15
    fp_style: gaussian
    use_clusters: false
    fp_params:
    keywords: P B3LYP/6-311+g(d) force
    nproc: 14
    charge: 0

train:
command: dp
machine:
batch_type: Shell
context_type: local
local_root: ./
remote_root: /home/sekinelabo/Documents/DeepMD-Sodium/dpgen/
resources:
gpu_per_node: 1
group_size: 14
batch_type: Shell
model_devi:
command: lmp_mpi
machine:
batch_type: Shell
local_root: ./
remote_root: /home/sekinelabo/Documents/DeepMD-Sodium/dpgen/
context_type: local
resources:
gpu_per_node: 1
group_size: 14
batch_type: Shell
fp:
command: g16 < input > output
machine:
batch_type: Shell
local_root: ./
remote_root: /home/sekinelabo/Documents/DeepMD-Sodium/dpgen/
context_type: local
resources:
gpu_per_node: 1
cpu_per_node: 14
group_size: 1
batch_type: Shell

[njzjz]

Just appending ||: to a bash command will make the exit code zero, as shown below.

dpgen/examples/machine/DeePMD-kit-1.x/machine-pbs-gaussian.json

Line 86 in 042ea32

"command": "/public/home/tzhu/g16/g16 < input || :"

[paseri3739]

Thank you for your replying! That's solved for your answer!