Add files via upload

The Annotator seems to be working fine when return_attention_weight is set false

Scripts/TestAnnotatorUnit.py

@@ -0,0 +1,53 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Mon Sep  2 09:57:21 2019
+@author: Hesham El Abd
+@Description: Test the Annotator module 
+"""
+from SelfAttentionLangModel.Models import EncoderModels
+import tensorflow as tf
+# First Testing the Model without returning the attention weights
+# bild the model
+testModel=EncoderModels.Annotator(embedding_dim=8,
+                                vocabulary_size=100,
+                                sequence_length=25,
+                                num_encoder_layer=4,num_heads=4,
+                                num_neuron_pointwise=32,
+                                return_attent_weights=False,
+                                rate=0.1,   
+                                num_dense_units=1,
+                                dense_activation=None)
+testModel.compile(optimizer="adam",
+                  loss="mae")
+testTensor=tf.random.uniform(shape=(100,25),minval=0,maxval=82,
+                                 dtype= tf.dtypes.int32)
+testTensorTarget=tf.random.uniform((100,1))
+# call the model to check the forward pass is working:
+dumRes=testModel(testTensor,False)
+print("Forward pass returned with a Tensor that has the following shape: "+
+      str(dumRes.shape))
+# print the summary of the model
+testModel.summary()
+## Test Model Training:
+testModel.fit(x=testTensor,y=testTensorTarget,batch_size=10)
+## Evaluating the Models with return self attention weights:
+testModel2=EncoderModels.Modeler(embedding_dim=8,
+                                vocabulary_size=82,
+                                conditional_string_length=100,
+                                num_encoder_layer=1,num_heads=4,
+                                num_neuron_pointwise=32,
+                                return_attent_weights=True,
+                                rate=0.1)
+testModel2.compile(optimizer="adam",
+                  loss=tf.keras.losses.SparseCategoricalCrossentropy())
+
+dumRes2,attent_weight=testModel2(testTensor,False)
+print("Forward pass returned with an output Tensor that has the following shape: "+
+      str(dumRes2.shape)+ "\n and and a list of self-attention matrices, the length"+
+      "of the list is "+str(len(attent_weight))+" and the shape of the first matrix is"+
+      str(attent_weight[0].shape))
+# print the summary of the model
+testModel2.summary()
+## Test Model Training:
+testModel2.fit(x=testTensor,y=testTensor,batch_size=10)