Schema for Active Learning Configuration¶
The schema to configure the input file for active learning process.
Schema:¶
### Schema for active learning configuration file
train: ### ANCHOR: Training ML model
type: dict
required: True
schema:
init_data_paths: # List of paths to initial data.
type: list
required: True
preprocess_data: # Arguments for processing data
type: dict
schema:
trainset_ratio: # Ratio of training set. Default is 0.9
type: float
validset_ratio: # Ratio of validation set. Default is 0.1
type: float
num_cores: # number of cores for building graph data. Default is 1
type: integer
ase_kwargs: # Custom keywords to access Energy/Force/Stress from ASE extxyz files. Default is empty dict, meaning use default ASE keywords.
type: dict
default: {}
schema:
energy_key: # keyword for energy in ASE extxyz file. Default is 'energy'
type: string
force_key: # keyword for force in ASE extxyz file. Default is 'forces'
type: string
stress_key: # keyword for stress in ASE extxyz file. Default is 'stress'
type: string
continue_train: # continue checkpoints from previous iteration. Default is True
type: boolean
num_grad_updates: # Maximum number of updates to guess num_epochs. Default is None
type: integer
distributed:
type: dict
schema:
distributed_backend: # choices: 'mpi' or 'nccl' 'gloo'
type: string
cluster_type: # choices: 'slurm' or 'sge'
type: string
gpu_per_node: # only need in SGE batch type. Default is 1
type: integer
num_models: # Number of models to train. Default is 1
type: integer
mlp_model: # ML model type. Default is 'sevenn_mliap'. Choices: 'sevenn', 'sevenn_mliap'
type: string
default: 'sevenn_mliap'
allowed: ['sevenn_mliap', 'sevenn']
sevenn_args: ### Args for training Sevenn. See: https://github.com/MDIL-SNU/SevenNet/blob/main/example_inputs/training/input_full.yml
type: dict
schema:
model:
type: dict
train:
type: dict
data: # Only need parameters in this schema
type: dict
schema:
batch_size: # Per GPU batch size. E.g., 300
type: integer
shift: # One of 'per_atom_energy_mean*', 'elemwise_reference_energies', float
type: [string, float]
scale: # One of 'force_rms*', 'per_atom_energy_std', 'elemwise_force_rms', float
type: [string, float]
md: ### ANCHOR: Run MD exploration
type: dict
required: True
schema:
committee_std: # Committee standard deviation for each property
type: dict
required: True
schema:
e_std_lo: # Lower bound of committee standard deviation for energy. E.g., 0.05
type: float
default: 0.0
e_std_hi: # Upper bound of committee standard deviation for energy. E.g., 0.15
type: float
default: 0.15
f_std_lo: # Lower bound of committee standard deviation for force. E.g., 0.05
type: float
default: 0.0
f_std_hi: # Upper bound of committee standard deviation for force. E.g., 0.15
type: float
default: 0.15
s_std_lo: # Lower bound of committee standard deviation for stress. E.g., 0.05
type: float
default: 0.0
s_std_hi: # Upper bound of committee standard deviation for stress. Default is None, meaning do not use stress criterion.
type: float
compute_stress: # Whether to compute stress for stress criterion. Default is True
type: boolean
default: True
rel_force: # Whether to use relative force std (w.r.t. force magnitude) for force criterion. Default is None
type: float
rel_stress: # Whether to use relative stress std (w.r.t. stress magnitude) for stress criterion. Default is None
type: float
checkpoint_conversion: # Parameters for converting checkpoints to MD calculator format
type: dict
schema:
checkpoint_idx: # indices of models/checkpoints in the committee to be used in MD run.
type: integer
extra_kwargs: # additional keyword arguments for conversion functions, depending on `mlp_model`. E.g., 'use_cueq'. 'use_flash', 'modal', etc.
type: dict
schema:
use_cueq: # whether to use charge equilibration. Default is False
type: boolean
default: False
use_flash: # whether to use FLASH formalism. Default is False
type: boolean
default: False
modal: # modal type for Sevenn model. Default is None
type: string
# allowed: ['single', 'multi']
dftd3: # whether to add DFT-D3 correction. Default is None
type: dict
schema:
damping: # damping method. Default is "zero"
type: string
allowed: ['zero', 'bj']
xc: # exchange-correlation functional. Default is "pbe"
type: string
default: "pbe"
cutoff: # cutoff distance in Angstrom. Default value is 95bohr := 50 angstrom. Units 1 (Bohr radius) = 0.52917721 (Å)
type: float
default: 50.2
cn_cutoff: # coordination number cutoff distance in Angstrom. Default value is 40bohr := 21 angstrom.
type: float
default: 21.17
d3package: # D3 package to use. Choices: 'sevenn' or 'lammps'. Default is 'sevenn'
type: string
allowed: ['sevenn', 'lammps']
default: 'sevenn'
init_struct_paths: # List of paths of initial structures. Accept wildcard, e.g., "path/to/dir/*.extxyz"
type: list
required: True
md_calculator: # MD calculator. Default is 'ase'. Choices: 'ase', 'lammps'
type: string
allowed: ['ase', 'lammps']
common_md_args: # define common parameters for MD. Accept all keyword fors [ASE MD schema](https://thangckt.github.io/alff_doc/schema/config_ase/)
type: dict # These common parameters will be overwritten by the parameters in `sampling_spaces`.
allow_unknown: True
schema:
dt: # timestep in fs. Default is 1.0 fs
type: float
thermostat: # define common thermostat. default is 'nose_hoover_chain'
type: string
barostat: # define common barostat. default is 'iso_nose_hoover_chain'
type: string
sampling_spaces: ### list of dicts of sampling spaces for MD run. For each dict:
# If set both 'stresses' and 'temperatures', it will use NPT ensemble.
# If set only 'temperatures', it will use NVT ensemble.
# If none of them, it will use NVE ensemble.
type: list
required: True
schema:
type: dict
allow_unknown: True
schema:
init_struct_idxs: # list of index of `init_struct_paths`, e.g., [0, 1, 2, 3]. Accept inputs as strings-of-range [1, 2, "6-10", "3-5:2"], which last item in ranges is inlusive.
type: list
temps: # list of temperatures in K, e.g., [1, 300, 1500]
type: list
pressures: # list of external stresses in GPa. Default is None if not set. Accept float, 6-vector, or 3x3 matrix.
type: [float, list]
# Other arguments as keywords for [ASE MD schema](https://thangckt.github.io/alff_doc/schema/config_ase/), except these keys: 'temperature', 'stress'.
dft: ### ANCHOR: DFT calculators
type: dict
required: True
schema:
calc_args: # accept all keywords for ['ase.calc'](https://thangckt.github.io/alff_doc/schema/config_ase/)
type: dict
required: True
schema:
gpaw: # GPAW calculator parameters
type: dict
required: True
dftd3: # DFT-D3 calculator for Van der Waals correction
type: dict
exclude_result_dftd3: # if True, exclude DFTD3 energy/force/stress from the final results. Default is False
type: boolean
default: False
Example config 1:¶
### Example configuration file for active learning with ALFF
#####ANCHOR Training
train:
init_data_paths:
- ../1_gendata/*/*/02_gendata/data_label.extxyz
- ../1_gendata/1_iteration_data
preprocess_data:
trainset_ratio: 0.9
validset_ratio: 0.1
num_cores: 1 # number of cores for building graph data
# ase_kwargs:
# energy_key: 'ref_energy' # keyword for energy in ASE extxyz file. Default is 'energy'
# force_key: 'ref_forces' # keyword for force in ASE extxyz file. Default is 'forces'
# stress_key: 'ref_stress' # keyword for stress in ASE extxyz file. Default is 'stress'
continue_train: True # continue checkpoints from previous iteration. Default is True
num_grad_updates: 100000 # Maximum number of updates to guess num_epochs. Default is None
distributed:
distributed_backend: 'nccl' # choices: 'mpi' or 'nccl' 'gloo'
cluster_type: 'slurm' # choices: 'slurm' or 'sge'
# gpu_per_node: 1 # only need in sge
num_models: 4
mlp_model: sevenn # 'sevenn_mliap', 'sevenn',
sevenn_args: # Updated: Dec 17, 2024. See: https://github.com/MDIL-SNU/SevenNet/blob/main/example_inputs/training/input_full.yaml
model:
chemical_species: ['Mo', 'W', 'S', 'Se', 'Te'] # Elements model should know. [ 'Univ' | 'Auto' | manual_user_input ]
cutoff: 5.0 # Cutoff radius in Angstroms. If two atoms are within the cutoff, they are connected.
channel: 32 # The multiplicity(channel) of node features.
lmax: 2 # Maximum order of irreducible representations (rotation order).
num_convolution_layer: 4 # The number of message passing layers.
# irreps_manual: # Manually set irreps of the model in each layer (e.g., 128 channels + 5 layers)
#- "128x0e"
#- "128x0e+64x1e+32x2e"
#- "128x0e+64x1e+32x2e"
#- "128x0e+64x1e+32x2e"
#- "128x0e+64x1e+32x2e"
#- "128x0e"
weight_nn_hidden_neurons: [64, 64] # Hidden neurons in convolution weight neural network
radial_basis: # Function and its parameters to encode radial distance
radial_basis_name: 'bessel' # Only 'bessel' is currently supported
bessel_basis_num: 8
cutoff_function: # Envelop function, multiplied to radial_basis functions to init edge featrues
cutoff_function_name: 'poly_cut' # {'poly_cut' and 'poly_cut_p_value'} or {'XPLOR' and 'cutoff_on'}
poly_cut_p_value: 6
act_gate: {'e': 'silu', 'o': 'tanh'} # Equivalent to 'nonlinearity_gates' in nequip
act_scalar: {'e': 'silu', 'o': 'tanh'} # Equivalent to 'nonlinearity_scalars' in nequip
is_parity: False # Pairy True (E(3) group) or False (to SE(3) group)
self_connection_type: linear # Default is 'nequip'. 'linear' is used for SevenNet-0.
interaction_type: nequip
conv_denominator: "avg_num_neigh" # Valid options are "avg_num_neigh*", "sqrt_avg_num_neigh", or float
train_denominator: False # Enable training for denominator in convolution layer
train_shift_scale: False # Enable training for shift & scale in output layer
train:
random_seed: 1
train_shuffle: True
is_train_stress: True # Includes stress in the loss function
epoch: 3 # Ends training after this number of epochs
per_epoch: 20 # Generate checkpoints every this epoch
# loss: 'Huber' # Default is 'mse' (mean squared error)
# loss_param:
# delta: 0.01
# Each optimizer and scheduler have different available parameters.
# You can refer to sevenn/train/optim.py for supporting optimizer & schedulers
optimizer: 'adam' # Options available are 'sgd', 'adagrad', 'adam', 'adamw', 'radam'
optim_param:
lr: 5.0e-4
scheduler: linearlr
scheduler_param:
start_factor: 1.0
total_iters: 3 # {..epoch}
end_factor: 1.0e-7
# scheduler: 'reducelronplateau' # One of 'steplr', 'multisteplr', 'exponentiallr', 'cosineannealinglr', 'reducelronplateau', 'linearlr'
# scheduler_param:
# factor: 0.75
# patience: 2
# threshold: 5.0e-5 # only changes large than this value will be considered as a change
# min_lr: 1.0e-12 # minimum learning rate
# scheduler: exponentiallr
# scheduler_param:
# gamma: 0.95 # large gamma means slower decay
force_loss_weight: 1.0 # Coefficient for force loss
stress_loss_weight: 1.0e-4 # 1.0e-3 1.0e-6 # Coefficient for stress loss (to kbar unit), kbar = 0.1 GPa
# ['target y', 'metric']
# Target y: TotalEnergy, Energy, Force, Stress, Stress_GPa, TotalLoss
# Metric : RMSE, MAE, or Loss
error_record:
- ['Energy', 'RMSE']
- ['Force', 'RMSE']
- ['Stress', 'RMSE']
# - ['Stress_GPa', 'RMSE']
- ['Energy', 'Loss']
- ['Force', 'Loss']
- ['Stress', 'Loss']
- ['TotalLoss', 'None']
best_metric: TotalLoss
### THANG: do not use this, just set `init_checkpoints` above
# Continue training model from given checkpoint, or pre-trained model checkpoint for fine-tuning
#continue:
#checkpoint: 'checkpoint_best.pth' # Checkpoint of pre-trained model or a model want to continue training.
#reset_optimizer: False # Set True for fine-tuning
#reset_scheduler: False # Set True for fine-tuning
data:
batch_size: 280 # 250 # Per GPU batch size.
shift: 'per_atom_energy_mean' # One of 'per_atom_energy_mean*', 'elemwise_reference_energies', float
scale: 'force_rms' # One of 'force_rms*', 'per_atom_energy_std', 'elemwise_force_rms', float
##### ANCHOR: Run MD
md:
committee_std:
e_std_lo: 0.05
e_std_hi: 0.15
f_std_lo: 0.05
f_std_hi: 0.15
s_std_lo: 0.05
s_std_hi: 0.15
checkpoint_conversion:
checkpoint_idx: 2 # index of models used for MD
dftd3: # add DFT-D3 correction during MD simulation
damping: 'zero' # damping method. Choices: "damp_zero", "damp_bj". Default is "damp_zero"
xc: 'pbe' # exchange-correlation functional. Default is "pbe"
md_calculator: 'lammps' # 'lammps', 'ase'
common_md_args:
dt: 0.001 # unit ASE: 1 fs; LAMMPS: 0.001 ps
thermostat: 'nose_hoover_chain' # 'langevin', 'nose_hoover_chain'. Nose-Hoover chain only for LAMMPS.
barostat: 'nose_hoover_chain' # ASE:'parrinello_rahman'. LAMMPS: 'nose_hoover_chain'
tdamp: 100
pdamp: 1000
init_struct_paths:
### MoX2 & WX2 bulk
- 0init_struct/bulk_MoX2_2x2x1/MoS2_mx2_2H_02x02x01 # '0-5' bulk MoX2
- 0init_struct/bulk_MoX2_2x2x1/MoS2_mx2_1T_02x02x01
- 0init_struct/bulk_MoX2_2x2x1/MoSe2_mx2_2H_02x02x01
- 0init_struct/bulk_MoX2_2x2x1/MoSe2_mx2_1T_02x02x01
- 0init_struct/bulk_MoX2_2x2x1/MoTe2_mx2_2H_02x02x01
- 0init_struct/bulk_MoX2_2x2x1/MoTe2_mx2_1T_02x02x01 # 5
- 0init_struct/bulk_WX2_2x2x1/WS2_mx2_2H_02x02x01 # '6-11' bulk WX2
- 0init_struct/bulk_WX2_2x2x1/WS2_mx2_1T_02x02x01
- 0init_struct/bulk_WX2_2x2x1/WSe2_mx2_2H_02x02x01
- 0init_struct/bulk_WX2_2x2x1/WSe2_mx2_1T_02x02x01
- 0init_struct/bulk_WX2_2x2x1/WTe2_mx2_2H_02x02x01
- 0init_struct/bulk_WX2_2x2x1/WTe2_mx2_1T_02x02x01 # 11
### MoX2 & WX2 layer
- 0init_struct/layer_MoX2_2x2x1/MoS2_mx2_2H_02x02x01 # '24-29' layer MoX2
- 0init_struct/layer_MoX2_2x2x1/MoS2_mx2_1T_02x02x01
- 0init_struct/layer_MoX2_2x2x1/MoSe2_mx2_2H_02x02x01
- 0init_struct/layer_MoX2_2x2x1/MoSe2_mx2_1T_02x02x01
- 0init_struct/layer_MoX2_2x2x1/MoTe2_mx2_2H_02x02x01
- 0init_struct/layer_MoX2_2x2x1/MoTe2_mx2_1T_02x02x01 # 29
- 0init_struct/layer_WX2_2x2x1/WS2_mx2_2H_02x02x01 # '30-35' layer WX2
- 0init_struct/layer_WX2_2x2x1/WS2_mx2_1T_02x02x01
- 0init_struct/layer_WX2_2x2x1/WSe2_mx2_2H_02x02x01
- 0init_struct/layer_WX2_2x2x1/WSe2_mx2_1T_02x02x01
- 0init_struct/layer_WX2_2x2x1/WTe2_mx2_2H_02x02x01
- 0init_struct/layer_WX2_2x2x1/WTe2_mx2_1T_02x02x01 # 35
sampling_spaces:
### If set both stress and temperature, it will use NPT ensemble.
### If set only temperature, it will use NVT ensemble.
### If none of them, it will use NVE ensemble.
- {}
- {}
### MoX2 & WX2
- init_struct_idxs: ['0-5','6-11','12-17','18-23', '24-29','30-35'] ##TODO BULK & LAYER: NVT
equil_steps: 20000
traj_freq: 5
num_frames: 20
temps: [1, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, 1100, 1200] # temperatures in K
- init_struct_idxs: ['0-5','6-11','12-17','18-23', '24-29','30-35'] ##TODO BULK & LAYER: NPT xy
equil_steps: 20000
traj_freq: 5
num_frames: 20
temps: [1, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, 1100, 1200] # temperatures in K
pressures: [0, -0.5] # stresses in GPa
mask: [1,1,0] # disable z-direction barostat
- init_struct_idxs: ['0-5','6-11','12-17','18-23'] ##TODO BULK: NPT xyz
equil_steps: 20000
traj_freq: 5
num_frames: 20
temps: [1, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, 1100, 1200]
pressures: [0, -0.5, -0.3, -0.1, 0.1, 0.3, 0.5] # stresses in GPa
mask: [1,1,1]
deform_limit: 0.9
##### ANCHOR: DFT calculation
dft:
calc_args:
gpaw: ### accept GPAW parameters
mode:
name: 'pw' # use PlaneWave method energy cutoff in eV
ecut: 500
xc: "PBE" # exchange-correlation functional
kpts:
density: 6
gamma: False # if not set `kpts`, then only Gamma-point is used
parallel:
sl_auto: True # enable ScaLAPACK parallelization
use_elpa: True # enable Elpa eigensolver
# augment_grids: True # use all cores for XC/Poisson solver
# gpu: True # enable GPU acceleration
dftd3: ### DFT-D3 method for Van der Waals correction
damping: "d3zero" # use DFT-D3 damping. Default is "d3zero" (zero-damping). Choices: "d3bj","d3zero","d3bjm","d3zerom","d3op".