| /** |
| * Copyright (c) 2020 CEA LIST |
| * |
| * All rights reserved. This program and the accompanying materials are |
| * made available under the terms of the Eclipse License 2.0 which |
| * accompanies this distribution, and is available at |
| * https://www.eclipse.org/legal/epl-2.0/ |
| * |
| * SPDX-License-Identifier: EPL-2.0 |
| * |
| * Contributors: |
| * Abhishek Djeachandrane - Initial API and implementation |
| */ |
| package org.eclipse.papyrus.aiml.gen.keras; |
| |
| import java.util.ArrayList; |
| import java.util.List; |
| import org.eclipse.emf.common.util.EList; |
| import org.eclipse.papyrus.aiml.gen.keras.GenCheck; |
| import org.eclipse.papyrus.aiml.gen.keras.ModelOps; |
| import org.eclipse.papyrus.aiml.gen.keras.ModuleInfo; |
| import org.eclipse.papyrus.aiml.gen.keras.ModuleInfoAttributes; |
| import org.eclipse.papyrus.aiml.gen.keras.RecurrentLayerOps; |
| import org.eclipse.papyrus.aiml.profile.AIML.Convolution_layers.Convolution_layers; |
| import org.eclipse.papyrus.aiml.profile.AIML.Linear_layers.Linear; |
| import org.eclipse.papyrus.aiml.profile.AIML.Linear_layers.Linear_layers; |
| import org.eclipse.papyrus.aiml.profile.AIML.Module.Model; |
| import org.eclipse.papyrus.aiml.profile.AIML.Recurrent_layers.GRU; |
| import org.eclipse.papyrus.aiml.profile.AIML.Recurrent_layers.LSTM; |
| import org.eclipse.papyrus.aiml.profile.AIML.Recurrent_layers.Recurrent_layers; |
| import org.eclipse.papyrus.aiml.profile.AIML.Tensor.Tensor; |
| import org.eclipse.uml2.uml.Connector; |
| import org.eclipse.uml2.uml.Property; |
| import org.eclipse.uml2.uml.Type; |
| import org.eclipse.uml2.uml.util.UMLUtil; |
| import org.eclipse.xtend2.lib.StringConcatenation; |
| import org.eclipse.xtext.xbase.lib.Conversions; |
| import org.eclipse.xtext.xbase.lib.IterableExtensions; |
| import org.eclipse.xtext.xbase.lib.StringExtensions; |
| |
| @SuppressWarnings("all") |
| public class GenKeras { |
| public static CharSequence genKeras(final org.eclipse.uml2.uml.Class clazz) { |
| final Model module = UMLUtil.<Model>getStereotypeApplication(clazz, Model.class); |
| if ((module != null)) { |
| return GenKeras.genModule(module); |
| } else { |
| return ""; |
| } |
| } |
| |
| /** |
| * Generate a model with different elements necessary for the compilation |
| */ |
| public static CharSequence genModule(final Model module) { |
| StringConcatenation _builder = new StringConcatenation(); |
| CharSequence _genDefaultTemplate = GenKeras.genDefaultTemplate(module); |
| _builder.append(_genDefaultTemplate); |
| _builder.newLineIfNotEmpty(); |
| _builder.newLine(); |
| CharSequence _genModuleInit = GenKeras.genModuleInit(module); |
| _builder.append(_genModuleInit); |
| _builder.newLineIfNotEmpty(); |
| _builder.newLine(); |
| _builder.append("\t"); |
| CharSequence _genModuleCall = GenKeras.genModuleCall(module); |
| _builder.append(_genModuleCall, "\t"); |
| _builder.newLineIfNotEmpty(); |
| _builder.newLine(); |
| CharSequence _genModuleCompilation = GenKeras.genModuleCompilation(module); |
| _builder.append(_genModuleCompilation); |
| _builder.newLineIfNotEmpty(); |
| _builder.newLine(); |
| _builder.newLine(); |
| _builder.newLine(); |
| _builder.newLine(); |
| _builder.newLine(); |
| return _builder; |
| } |
| |
| /** |
| * Generate the model instance and compilation |
| */ |
| public static CharSequence genModuleCompilation(final Model module) { |
| StringConcatenation _builder = new StringConcatenation(); |
| String _firstLower = StringExtensions.toFirstLower(module.getBase_Class().getName()); |
| _builder.append(_firstLower); |
| _builder.append(" = "); |
| String _name = module.getBase_Class().getName(); |
| _builder.append(_name); |
| _builder.append("()"); |
| _builder.newLineIfNotEmpty(); |
| _builder.append("\t\t"); |
| ModuleInfoAttributes study = ModelOps.modelAttributesToCollectProperly(module); |
| _builder.newLineIfNotEmpty(); |
| List<String> _list = IterableExtensions.<String>toList(study.attributesName); |
| _builder.append(_list); |
| _builder.newLineIfNotEmpty(); |
| List<Object> _list_1 = IterableExtensions.<Object>toList(study.attributesValue); |
| _builder.append(_list_1); |
| _builder.newLineIfNotEmpty(); |
| List<String> _list_2 = IterableExtensions.<String>toList(study.attributesType); |
| _builder.append(_list_2); |
| _builder.newLineIfNotEmpty(); |
| { |
| List<Object> _list_3 = IterableExtensions.<Object>toList(study.attributesValue); |
| for(final Object n : _list_3) { |
| _builder.newLine(); |
| } |
| } |
| String study2 = ModelOps.modelAttributesToDisplayForCompilationProperly(module); |
| _builder.newLineIfNotEmpty(); |
| _builder.append(study2); |
| _builder.newLineIfNotEmpty(); |
| return _builder; |
| } |
| |
| /** |
| * Generate the basic template with importation in the head of the code |
| */ |
| public static CharSequence genDefaultTemplate(final Model module) { |
| StringConcatenation _builder = new StringConcatenation(); |
| _builder.append("# generated by Papyrus from module "); |
| Class<? extends org.eclipse.uml2.uml.Class> _class = module.getBase_Class().getClass(); |
| _builder.append(_class); |
| _builder.newLineIfNotEmpty(); |
| _builder.newLine(); |
| _builder.append("import tensorflow as tf"); |
| _builder.newLine(); |
| _builder.append("from tensorflow import keras"); |
| _builder.newLine(); |
| _builder.append("inputs=[(1,2),(3,4)]"); |
| _builder.newLine(); |
| return _builder; |
| } |
| |
| /** |
| * Generate layer instance through modeling |
| */ |
| public static CharSequence genModuleInit(final Model module) { |
| StringConcatenation _builder = new StringConcatenation(); |
| _builder.append("class "); |
| String _name = module.getBase_Class().getName(); |
| _builder.append(_name); |
| _builder.append("(tf.keras.Model):"); |
| _builder.newLineIfNotEmpty(); |
| _builder.append("\t"); |
| _builder.append("def __init__(self):"); |
| _builder.newLine(); |
| _builder.append("\t\t"); |
| _builder.append("super(MyModel, self).__init__()"); |
| _builder.newLine(); |
| _builder.append("\t\t"); |
| _builder.newLine(); |
| { |
| EList<Property> _ownedAttributes = module.getBase_Class().getOwnedAttributes(); |
| for(final Property layer : _ownedAttributes) { |
| _builder.append("\t\t"); |
| final Type type = layer.getType(); |
| _builder.newLineIfNotEmpty(); |
| _builder.append("\t\t"); |
| final Model layerSt = UMLUtil.<Model>getStereotypeApplication(type, Model.class); |
| _builder.newLineIfNotEmpty(); |
| _builder.append("\t\t"); |
| CharSequence _genLayer = GenKeras.genLayer(layerSt); |
| _builder.append(_genLayer, "\t\t"); |
| _builder.newLineIfNotEmpty(); |
| } |
| } |
| { |
| EList<Connector> _ownedConnectors = module.getBase_Class().getOwnedConnectors(); |
| for(final Connector connector : _ownedConnectors) { |
| } |
| } |
| return _builder; |
| } |
| |
| /** |
| * Generate the code which consist to determine the data flow between layers |
| * @param l the model ... |
| */ |
| public static CharSequence genModuleCall(final Model module) { |
| StringConcatenation _builder = new StringConcatenation(); |
| _builder.append("def call(self, inputs):"); |
| _builder.newLine(); |
| _builder.newLine(); |
| int nbr_lay = GenCheck.findNbrOfLayer(module); |
| _builder.newLineIfNotEmpty(); |
| Object unsortedInformationFlowList = ModelOps.genInputOutputLayerArrayWithFlow(Integer.valueOf(nbr_lay), module); |
| _builder.newLineIfNotEmpty(); |
| _builder.newLine(); |
| { |
| if ((unsortedInformationFlowList instanceof ModuleInfo)) { |
| ArrayList<String> sortedInformationFlowList = ModelOps.genFlowOrder(Integer.valueOf(nbr_lay), ((ModuleInfo)unsortedInformationFlowList)); |
| _builder.newLineIfNotEmpty(); |
| { |
| List<String> _list = IterableExtensions.<String>toList(sortedInformationFlowList); |
| for(final String layer : _list) { |
| _builder.append("\t"); |
| int _indexOf = IterableExtensions.<String>toList(sortedInformationFlowList).indexOf(layer); |
| int inputs = (_indexOf - 1); |
| _builder.newLineIfNotEmpty(); |
| { |
| if ((inputs < 0)) { |
| _builder.append("\t"); |
| _builder.append("x"); |
| int _indexOf_1 = IterableExtensions.<String>toList(sortedInformationFlowList).indexOf(layer); |
| _builder.append(_indexOf_1, "\t"); |
| _builder.append(" = self."); |
| _builder.append(layer, "\t"); |
| _builder.append("(input)"); |
| _builder.newLineIfNotEmpty(); |
| } else { |
| if (((inputs >= 0) && (inputs < (((Object[])Conversions.unwrapArray(sortedInformationFlowList, Object.class)).length - 2)))) { |
| _builder.append("\t"); |
| _builder.append("x"); |
| int _indexOf_2 = IterableExtensions.<String>toList(sortedInformationFlowList).indexOf(layer); |
| _builder.append(_indexOf_2, "\t"); |
| _builder.append(" = self."); |
| _builder.append(layer, "\t"); |
| _builder.append("(x"); |
| _builder.append(inputs, "\t"); |
| _builder.append(")"); |
| _builder.newLineIfNotEmpty(); |
| } else { |
| final ArrayList<String> _converted_sortedInformationFlowList = (ArrayList<String>)sortedInformationFlowList; |
| int _length = ((Object[])Conversions.unwrapArray(_converted_sortedInformationFlowList, Object.class)).length; |
| int _minus = (_length - 2); |
| boolean _equals = (inputs == _minus); |
| if (_equals) { |
| _builder.append("\t"); |
| _builder.append("output = self."); |
| _builder.append(layer, "\t"); |
| _builder.append("(x"); |
| _builder.append(inputs, "\t"); |
| _builder.append(")"); |
| _builder.newLineIfNotEmpty(); |
| } |
| } |
| } |
| } |
| } |
| } |
| } |
| } |
| _builder.append("\t"); |
| _builder.append("return output\t\t"); |
| _builder.newLine(); |
| return _builder; |
| } |
| |
| /** |
| * Generate each layer according to the model |
| */ |
| public static CharSequence genLayer(final Model layer) { |
| CharSequence _xifexpression = null; |
| if ((layer instanceof Convolution_layers)) { |
| _xifexpression = GenKeras.genConvolutionLayer(((Convolution_layers) layer)); |
| } else { |
| CharSequence _xifexpression_1 = null; |
| if ((layer instanceof Recurrent_layers)) { |
| _xifexpression_1 = GenKeras.genRecurrentLayer(((Recurrent_layers) layer)); |
| } else { |
| CharSequence _xifexpression_2 = null; |
| if ((layer instanceof Linear_layers)) { |
| _xifexpression_2 = GenKeras.genLinearLayer(((Linear_layers) layer)); |
| } |
| _xifexpression_1 = _xifexpression_2; |
| } |
| _xifexpression = _xifexpression_1; |
| } |
| return _xifexpression; |
| } |
| |
| /** |
| * Generate each convolution layer according to the model |
| */ |
| public static CharSequence genConvolutionLayer(final Convolution_layers convolutionLayer) { |
| StringConcatenation _builder = new StringConcatenation(); |
| return _builder; |
| } |
| |
| /** |
| * Generate each linear layer according to the model |
| */ |
| public static CharSequence genLinearLayer(final Linear_layers linearLayer) { |
| CharSequence _xifexpression = null; |
| if ((linearLayer instanceof Linear)) { |
| _xifexpression = GenKeras.genDense(((Linear) linearLayer)); |
| } |
| return _xifexpression; |
| } |
| |
| /** |
| * Generate each recurrent layer according to the model |
| */ |
| public static CharSequence genRecurrentLayer(final Recurrent_layers recurrentLayer) { |
| CharSequence _xifexpression = null; |
| if ((recurrentLayer instanceof LSTM)) { |
| _xifexpression = GenKeras.genLstm(((LSTM) recurrentLayer)); |
| } else { |
| CharSequence _xifexpression_1 = null; |
| if ((recurrentLayer instanceof GRU)) { |
| _xifexpression_1 = GenKeras.genGru(((GRU) recurrentLayer)); |
| } |
| _xifexpression = _xifexpression_1; |
| } |
| return _xifexpression; |
| } |
| |
| /** |
| * Generate each dense layer according to the model |
| */ |
| public static CharSequence genDense(final Linear denseLayer) { |
| StringConcatenation _builder = new StringConcatenation(); |
| _builder.append("self."); |
| String _name = denseLayer.getBase_Class().getName(); |
| _builder.append(_name); |
| _builder.append("=tf.keras.layers.Dense("); |
| int _in_features = denseLayer.getIn_features(); |
| _builder.append(_in_features); |
| _builder.append(")"); |
| _builder.newLineIfNotEmpty(); |
| return _builder; |
| } |
| |
| /** |
| * Generate each LSTM layer according to the model |
| */ |
| public static CharSequence genLstm(final LSTM lstmLayer) { |
| StringConcatenation _builder = new StringConcatenation(); |
| _builder.append("\t\t\t\t\t\t\t\t\t\t\t\t"); |
| String tt = RecurrentLayerOps.lstmAttributesToDisplay(lstmLayer); |
| _builder.newLineIfNotEmpty(); |
| _builder.append(tt); |
| _builder.newLineIfNotEmpty(); |
| return _builder; |
| } |
| |
| /** |
| * Generate each GRU layer according to the model |
| */ |
| public static CharSequence genGru(final GRU gruLayer) { |
| StringConcatenation _builder = new StringConcatenation(); |
| _builder.append("\t\t\t\t"); |
| String tt = RecurrentLayerOps.lstmAttributesToDisplay(gruLayer); |
| _builder.newLineIfNotEmpty(); |
| _builder.append(tt); |
| _builder.newLineIfNotEmpty(); |
| return _builder; |
| } |
| |
| /** |
| * Generate each tensor according to the model |
| */ |
| public static CharSequence genTensor(final Tensor tensor) { |
| StringConcatenation _builder = new StringConcatenation(); |
| _builder.append("tensor ..."); |
| _builder.newLine(); |
| return _builder; |
| } |
| } |
