deepmd.pd.loss

deepmd.pd.loss#

Submodules#

Classes#

`EnergyHessianStdLoss`	Helper class that provides a standard way to create an ABC using
`EnergyStdLoss`	Helper class that provides a standard way to create an ABC using
`TaskLoss`	Helper class that provides a standard way to create an ABC using

Package Contents#

class deepmd.pd.loss.EnergyHessianStdLoss(start_pref_h=0.0, limit_pref_h=0.0, **kwargs)[source]#

Bases: EnergyStdLoss

Helper class that provides a standard way to create an ABC using inheritance.

has_h#

start_pref_h = 0.0#

limit_pref_h = 0.0#

forward(input_dict, model, label, natoms, learning_rate, mae=False)[source]#

Return loss on energy and force.

Parameters:

input_dictdict[str, paddle.Tensor]: Model inputs.
modelpaddle.nn.Layer: Model to be used to output the predictions.
labeldict[str, paddle.Tensor]: Labels.
natomsint: The local atom number.

Returns:

model_pred: dict[str, paddle.Tensor]: Model predictions.
loss: paddle.Tensor: Loss for model to minimize.
more_loss: dict[str, paddle.Tensor]: Other losses for display.

property label_requirement: list[deepmd.utils.data.DataRequirementItem]#: Add hessian label requirement needed for this loss calculation.

class deepmd.pd.loss.EnergyStdLoss(starter_learning_rate=1.0, start_pref_e=0.0, limit_pref_e=0.0, start_pref_f=0.0, limit_pref_f=0.0, start_pref_v=0.0, limit_pref_v=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_l1_all: bool = False, inference=False, use_huber=False, huber_delta=0.01, **kwargs)[source]#

Bases: deepmd.pd.loss.loss.TaskLoss

Helper class that provides a standard way to create an ABC using inheritance.

starter_learning_rate = 1.0#

has_e#

has_f#

has_v#

has_ae#

has_pf#

has_gf#

start_pref_e = 0.0#

limit_pref_e = 0.0#

start_pref_f = 0.0#

limit_pref_f = 0.0#

start_pref_v = 0.0#

limit_pref_v = 0.0#

start_pref_ae = 0.0#

limit_pref_ae = 0.0#

start_pref_pf = 0.0#

limit_pref_pf = 0.0#

start_pref_gf = 0.0#

limit_pref_gf = 0.0#

relative_f = None#

enable_atom_ener_coeff = False#

numb_generalized_coord = 0#

use_l1_all = False#

inference = False#

use_huber = False#

huber_delta = 0.01#

forward(input_dict, model, label, natoms, learning_rate, mae=False)[source]#

Return loss on energy and force.

Parameters:

input_dictdict[str, paddle.Tensor]: Model inputs.
modelpaddle.nn.Layer: Model to be used to output the predictions.
labeldict[str, paddle.Tensor]: Labels.
natomsint: The local atom number.

Returns:

model_pred: dict[str, paddle.Tensor]: Model predictions.
loss: paddle.Tensor: Loss for model to minimize.
more_loss: dict[str, paddle.Tensor]: Other losses for display.

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

serialize() → dict[source]#

Serialize the loss module.

Returns:

dict: The serialized loss module

classmethod deserialize(data: dict) → deepmd.pd.loss.loss.TaskLoss[source]#

Deserialize the loss module.

Parameters:

datadict: The serialized loss module

Returns:

Loss: The deserialized loss module

class deepmd.pd.loss.TaskLoss(**kwargs)[source]#

Bases: paddle.nn.Layer, abc.ABC, make_plugin_registry('loss')

Helper class that provides a standard way to create an ABC using inheritance.

abstractmethod forward(input_dict, model, label, natoms, learning_rate)[source]#: Return loss .

property label_requirement: list[deepmd.utils.data.DataRequirementItem]#

Abstractmethod:

Return data label requirements needed for this loss calculation.

static display_if_exist(loss: paddle.Tensor, find_property: float) → paddle.Tensor[source]#

Display NaN if labeled property is not found.

Parameters:

losspaddle.Tensor: the loss tensor
find_propertyfloat: whether the property is found

classmethod get_loss(loss_params: dict) → TaskLoss[source]#

Get the loss module by the parameters.

By default, all the parameters are directly passed to the constructor. If not, override this method.

Parameters:

loss_paramsdict: The loss parameters

Returns:

TaskLoss: The loss module