deepmd.tf.loss.ener

deepmd.tf.loss.ener#

Classes#

`EnerStdLoss`	Standard loss function for DP models.
`EnerSpinLoss`	The abstract class for the loss function.
`EnerDipoleLoss`	The abstract class for the loss function.

Functions#

custom_huber_loss(predictions, targets[, delta])

Module Contents#

deepmd.tf.loss.ener.custom_huber_loss(predictions, targets, delta=1.0)[source]#

class deepmd.tf.loss.ener.EnerStdLoss(starter_learning_rate: float, start_pref_e: float = 0.02, limit_pref_e: float = 1.0, start_pref_f: float = 1000, limit_pref_f: float = 1.0, start_pref_v: float = 0.0, limit_pref_v: float = 0.0, start_pref_ae: float = 0.0, limit_pref_ae: float = 0.0, start_pref_pf: float = 0.0, limit_pref_pf: float = 0.0, relative_f: float | None = None, enable_atom_ener_coeff: bool = False, start_pref_gf: float = 0.0, limit_pref_gf: float = 0.0, numb_generalized_coord: int = 0, use_huber=False, huber_delta=0.01, **kwargs)[source]#

Bases: deepmd.tf.loss.loss.Loss

Standard loss function for DP models.

Parameters:

starter_learning_ratefloat: The learning rate at the start of the training.
start_pref_efloat: The prefactor of energy loss at the start of the training.
limit_pref_efloat: The prefactor of energy loss at the end of the training.
start_pref_ffloat: The prefactor of force loss at the start of the training.
limit_pref_ffloat: The prefactor of force loss at the end of the training.
start_pref_vfloat: The prefactor of virial loss at the start of the training.
limit_pref_vfloat: The prefactor of virial loss at the end of the training.
start_pref_aefloat: The prefactor of atomic energy loss at the start of the training.
limit_pref_aefloat: The prefactor of atomic energy loss at the end of the training.
start_pref_pffloat: The prefactor of atomic prefactor force loss at the start of the training.
limit_pref_pffloat: The prefactor of atomic prefactor force loss at the end of the training.
relative_ffloat: If provided, relative force error will be used in the loss. The difference of force will be normalized by the magnitude of the force in the label with a shift given by relative_f
enable_atom_ener_coeffbool: if true, the energy will be computed as sum_i c_i E_i
start_pref_gffloat: The prefactor of generalized force loss at the start of the training.
limit_pref_gffloat: The prefactor of generalized force loss at the end of the training.
numb_generalized_coordint: The dimension of generalized coordinates.
use_huberbool: Enables Huber loss calculation for energy/force/virial terms with user-defined threshold delta (D). The loss function smoothly transitions between L2 and L1 loss: - For absolute prediction errors within D: quadratic loss (0.5 * (error**2)) - For absolute errors exceeding D: linear loss (D * |error| - 0.5 * D) Formula: loss = 0.5 * (error**2) if |error| <= D else D * (|error| - 0.5 * D).
huber_deltafloat: The threshold delta (D) used for Huber loss, controlling transition between L2 and L1 loss.
**kwargs: Other keyword arguments.

starter_learning_rate[source]#

start_pref_e = 0.02[source]#

limit_pref_e = 1.0[source]#

start_pref_f = 1000[source]#

limit_pref_f = 1.0[source]#

start_pref_v = 0.0[source]#

limit_pref_v = 0.0[source]#

start_pref_ae = 0.0[source]#

limit_pref_ae = 0.0[source]#

start_pref_pf = 0.0[source]#

limit_pref_pf = 0.0[source]#

relative_f = None[source]#

enable_atom_ener_coeff = False[source]#

start_pref_gf = 0.0[source]#

limit_pref_gf = 0.0[source]#

numb_generalized_coord = 0[source]#

has_e[source]#

has_f[source]#

has_v[source]#

has_ae[source]#

has_pf[source]#

has_gf[source]#

use_huber = False[source]#

huber_delta = 0.01[source]#

build(learning_rate, natoms, model_dict, label_dict, suffix)[source]#

Build the loss function graph.

Parameters:

learning_ratetf.Tensor: learning rate
natomstf.Tensor: number of atoms
model_dictdict[str, tf.Tensor]: A dictionary that maps model keys to tensors
label_dictdict[str, tf.Tensor]: A dictionary that maps label keys to tensors
suffixstr: suffix

Returns:

tf.Tensor: the total squared loss
dict[str, tf.Tensor]: A dictionary that maps loss keys to more loss tensors

eval(sess, feed_dict, natoms)[source]#

Eval the loss function.

Parameters:

sesstf.Session: TensorFlow session
feed_dictdict[tf.placeholder, tf.Tensor]: A dictionary that maps graph elements to values
natomstf.Tensor: number of atoms

Returns:

dict: A dictionary that maps keys to values. It should contain key natoms

property label_requirement: list[deepmd.utils.data.DataRequirementItem][source]#: Return data label requirements needed for this loss calculation.

serialize(suffix: str = '') → dict[source]#

Serialize the loss module.

Parameters:

suffixstr: The suffix of the loss module

Returns:

dict: The serialized loss module

classmethod deserialize(data: dict, suffix: str = '') → deepmd.tf.loss.loss.Loss[source]#

Deserialize the loss module.

Parameters:

datadict: The serialized loss module
suffixstr: The suffix of the loss module

Returns:

Loss: The deserialized loss module

class deepmd.tf.loss.ener.EnerSpinLoss(starter_learning_rate: float, start_pref_e: float = 0.02, limit_pref_e: float = 1.0, start_pref_fr: float = 1000, limit_pref_fr: float = 1.0, start_pref_fm: float = 10000, limit_pref_fm: float = 10.0, start_pref_v: float = 0.0, limit_pref_v: float = 0.0, start_pref_ae: float = 0.0, limit_pref_ae: float = 0.0, start_pref_pf: float = 0.0, limit_pref_pf: float = 0.0, relative_f: float | None = None, enable_atom_ener_coeff: bool = False, use_spin: list | None = None)[source]#

Bases: deepmd.tf.loss.loss.Loss

The abstract class for the loss function.

starter_learning_rate[source]#

start_pref_e = 0.02[source]#

limit_pref_e = 1.0[source]#

start_pref_fr = 1000[source]#

limit_pref_fr = 1.0[source]#

start_pref_fm = 10000[source]#

limit_pref_fm = 10.0[source]#

start_pref_v = 0.0[source]#

limit_pref_v = 0.0[source]#

start_pref_ae = 0.0[source]#

limit_pref_ae = 0.0[source]#

start_pref_pf = 0.0[source]#

limit_pref_pf = 0.0[source]#

relative_f = None[source]#

enable_atom_ener_coeff = False[source]#

use_spin = None[source]#

has_e[source]#

has_fr[source]#

has_fm[source]#

has_v[source]#

has_ae[source]#

build(learning_rate, natoms, model_dict, label_dict, suffix)[source]#

Build the loss function graph.

Parameters:

learning_ratetf.Tensor: learning rate
natomstf.Tensor: number of atoms
model_dictdict[str, tf.Tensor]: A dictionary that maps model keys to tensors
label_dictdict[str, tf.Tensor]: A dictionary that maps label keys to tensors
suffixstr: suffix

Returns:

tf.Tensor: the total squared loss
dict[str, tf.Tensor]: A dictionary that maps loss keys to more loss tensors

eval(sess, feed_dict, natoms)[source]#

Eval the loss function.

Parameters:

sesstf.Session: TensorFlow session
feed_dictdict[tf.placeholder, tf.Tensor]: A dictionary that maps graph elements to values
natomstf.Tensor: number of atoms

Returns:

dict: A dictionary that maps keys to values. It should contain key natoms

print_header()[source]#

print_on_training(tb_writer, cur_batch, sess, natoms, feed_dict_test, feed_dict_batch)[source]#

property label_requirement: list[deepmd.utils.data.DataRequirementItem][source]#: Return data label requirements needed for this loss calculation.

class deepmd.tf.loss.ener.EnerDipoleLoss(starter_learning_rate: float, start_pref_e: float = 0.1, limit_pref_e: float = 1.0, start_pref_ed: float = 1.0, limit_pref_ed: float = 1.0)[source]#

Bases: deepmd.tf.loss.loss.Loss

The abstract class for the loss function.

starter_learning_rate[source]#

start_pref_e = 0.1[source]#

limit_pref_e = 1.0[source]#

start_pref_ed = 1.0[source]#

limit_pref_ed = 1.0[source]#

build(learning_rate, natoms, model_dict, label_dict, suffix)[source]#

Build the loss function graph.

Parameters:

learning_ratetf.Tensor: learning rate
natomstf.Tensor: number of atoms
model_dictdict[str, tf.Tensor]: A dictionary that maps model keys to tensors
label_dictdict[str, tf.Tensor]: A dictionary that maps label keys to tensors
suffixstr: suffix

Returns:

tf.Tensor: the total squared loss
dict[str, tf.Tensor]: A dictionary that maps loss keys to more loss tensors

eval(sess, feed_dict, natoms)[source]#

Eval the loss function.

Parameters:

sesstf.Session: TensorFlow session
feed_dictdict[tf.placeholder, tf.Tensor]: A dictionary that maps graph elements to values
natomstf.Tensor: number of atoms

Returns:

dict: A dictionary that maps keys to values. It should contain key natoms

property label_requirement: list[deepmd.utils.data.DataRequirementItem][source]#: Return data label requirements needed for this loss calculation.