MindElec

mindelec.architecture

API Name

Description

Supported Platforms

mindelec.architecture.FCSequential

The dense layer sequential.

Ascend

mindelec.architecture.InputScaleNet

Scale the input value to specified region.

Ascend

mindelec.architecture.LinearBlock

The LinearBlock.

Ascend

mindelec.architecture.MTLWeightedLossCell

The MTL strategy weighted multi-task losses automatically.

Ascend

mindelec.architecture.MultiScaleFCCell

The multi-scale network.

Ascend

mindelec.architecture.ResBlock

The ResBlock of dense layer.

Ascend

mindelec.architecture.get_activation

Gets the activation function.

Ascend

mindelec.common

API Name

Description

Supported Platforms

mindelec.common.L2

Calculates l2 metric.

Ascend

mindelec.common.LearningRate

Warmup-decay learning rate.

Ascend

mindelec.common.get_poly_lr

generate learning rate array

Ascend

mindelec.data

API Name

Description

Supported Platforms

mindelec.data.BBoxType

Bounding box for sampling space, only supports cube-shape sampling space, at present supports STATIC(0) and DYNAMIC(1).

Ascend

mindelec.data.BoundaryBC

Sampling data of boundary condition.

Ascend

mindelec.data.BoundaryIC

Sampling data of initial condition.

Ascend

mindelec.data.Dataset

Combine datasets together.

Ascend

mindelec.data.Equation

Sampling data of equation domain.

Ascend

mindelec.data.ExistedDataConfig

Set arguments of ExistedDataset.

Ascend

mindelec.data.ExistedDataset

Creates a dataset with given data path.

Ascend

mindelec.data.MaterialConfig

Material solution config for PointCloud-Tensor generation, which influence the material solving stage.

Ascend

mindelec.data.PointCloud

Read the stp files to generate PointCloud data, for downstream physical-equation AI simulation.

Ascend

mindelec.data.PointCloudSamplingConfig

Sampling space config for PointCloud-Tensor generation.

Ascend

mindelec.data.SamplingMode

Point sampling method, at present support UPPERBOUND(0) and DIMENSIONS(1).

Ascend

mindelec.data.StdPhysicalQuantity

Standard physical quantities fields that Maxwell equations concern about, material solving stage will deal with these standard physical fields.

Ascend

mindelec.geometry

API Name

Description

Supported Platforms

mindelec.geometry.create_config_from_edict

Convert from dict to SamplingConfig.

Ascend

mindelec.geometry.CSGDifference

CSG class for difference of geometry.

Ascend

mindelec.geometry.CSGIntersection

CSG class for intersection of geometries.

Ascend

mindelec.geometry.CSGUnion

CSG class for union of geometries.

Ascend

mindelec.geometry.CSGXOR

CSG class for xor of geometries.

Ascend

mindelec.geometry.Cuboid

Definition of Cuboid object.

Ascend

mindelec.geometry.Disk

Definition of Disk object.

Ascend

mindelec.geometry.Geometry

Definition of Geometry object.

Ascend

mindelec.geometry.GeometryWithTime

Definition of geometry with time.

Ascend

mindelec.geometry.HyperCube

Definition of HyperCube object.

Ascend

mindelec.geometry.Interval

Definition of Interval object.

Ascend

mindelec.geometry.PartSamplingConfig

Definition of partial sampling configuration.

Ascend

mindelec.geometry.Rectangle

Definition of Rectangle object.

Ascend

mindelec.geometry.SamplingConfig

Definition of global sampling configuration.

Ascend

mindelec.geometry.TimeDomain

Definition of Time Domain.

Ascend

mindelec.loss

API Name

Description

Supported Platforms

mindelec.loss.Constraints

Definition of the loss for all sub-dataset.

Ascend

mindelec.loss.NetWithEval

Encapsulation class of network with loss of eval.

Ascend

mindelec.loss.NetWithLoss

Encapsulation class of network with loss.

Ascend

mindelec.loss.get_loss_metric

Gets the loss function.

Ascend

mindelec.operators

API Name

Description

Supported Platforms

mindelec.operators.Grad

Computes and returns the gradients of the specified column of outputs with respect to the specified column of inputs.

Ascend

mindelec.operators.SecondOrderGrad

Computes and returns the second order gradients of the specified column of outputs with respect to the specified column of inputs.

Ascend

mindelec.solver

API Name

Description

Supported Platforms

mindelec.solver.EvalCallback

Evaluate the model during training.

Ascend

mindelec.solver.LossAndTimeMonitor

Monitor the loss in training.

Ascend

mindelec.solver.Problem

Base class of user-defined problems.

Ascend

mindelec.solver.Solver

High-Level API for training or inference.

Ascend

mindelec.vision

API Name

Description

Supported Platforms

mindelec.vision.MonitorEval

LossMonitor for eval.

Ascend

mindelec.vision.MonitorTrain

Loss monitor for train.

Ascend

mindelec.vision.image_to_video

Create video from existing images.

Ascend

mindelec.vision.plot_eh

Draw electric and magnetic field values of every timestep for 2D slices, and save them in path_image_save

Ascend

mindelec.vision.plot_s11

Draw s11-frequency curve and save it in path_image_save.

Ascend

mindelec.vision.print_graph_1d

Draw 1d scatter image

Ascend

mindelec.vision.print_graph_2d

Draw 2d scatter image

Ascend

mindelec.vision.vtk_structure

Generates 3D vtk file for visualizaiton.

Ascend