From 2523fdd829c0d740c8a9610ee96c5e1628ad6907 Mon Sep 17 00:00:00 2001
From: Ivan Kitanovski <ivan.kitanovski@gmail.com>
Date: Tue, 30 Apr 2019 13:46:32 +0200
Subject: [PATCH] Updated logging and gradlew permissons

---
 build.gradle                                  |  3 +--
 gradlew                                       |  0
 .../com/github/fairsearch/deltr/Trainer.java  | 20 +++++++++----------
 3 files changed, 10 insertions(+), 13 deletions(-)
 mode change 100644 => 100755 gradlew

diff --git a/build.gradle b/build.gradle
index e95dedd..ee51f9d 100644
--- a/build.gradle
+++ b/build.gradle
@@ -59,8 +59,7 @@ publishing {
         mavenJava(MavenPublication) {
             groupId 'com.github.fair-search'
             artifactId 'fairsearchdeltr-java'
-            version '1.0.0'
-
+            version '0.0.1'
             from components.java
         }
     }
diff --git a/gradlew b/gradlew
old mode 100644
new mode 100755
diff --git a/src/main/java/com/github/fairsearch/deltr/Trainer.java b/src/main/java/com/github/fairsearch/deltr/Trainer.java
index 5a2543d..f993e7d 100644
--- a/src/main/java/com/github/fairsearch/deltr/Trainer.java
+++ b/src/main/java/com/github/fairsearch/deltr/Trainer.java
@@ -91,14 +91,14 @@ public double[] train(int[] queryIds, int[] protectedElementFeature, INDArray fe
 
             });
 
-            LOGGER.info("pred: " + (System.currentTimeMillis() - stepStart));
+            LOGGER.info(String.format("Prediction computed in %d ms", (System.currentTimeMillis() - stepStart)));
             stepStart = System.currentTimeMillis();
 
             //get the cost/loss for all queries
             TrainStep trainStep = calculateCost(trainingScores, predictedScores, queryIds,
                     protectedElementFeature, dataPerQueryPredicted);
 
-            LOGGER.info(" cost: " + (System.currentTimeMillis() - stepStart));
+            LOGGER.info("Cost: " + (System.currentTimeMillis() - stepStart));
             stepStart = System.currentTimeMillis();
 
             INDArray J = trainStep.getCost().add(predictedScores.mul(predictedScores).mul(this.lambda));
@@ -107,7 +107,7 @@ public double[] train(int[] queryIds, int[] protectedElementFeature, INDArray fe
             INDArray grad = calculateGradient(featureMatrix, trainingScores, predictedScores, queryIds,
                     protectedElementFeature, dataPerQueryPredicted);
 
-            LOGGER.info(" grad: " + (System.currentTimeMillis() - stepStart));
+            LOGGER.info(String.format("Gradient computed in %d ms", (System.currentTimeMillis() - stepStart)));
 
             //add additional items in trainStep
             trainStep.setOmega(omega);
@@ -125,7 +125,7 @@ public double[] train(int[] queryIds, int[] protectedElementFeature, INDArray fe
             this.log.add(trainStep);
 
             // log iteration
-            LOGGER.info("Iteration-" + t +":" + (System.currentTimeMillis() - startTime));
+            LOGGER.info(String.format("Iteration %d done in %d ms", t, (System.currentTimeMillis() - startTime)));
 
         }
 
@@ -160,13 +160,13 @@ private INDArray calculateGradient(INDArray trainingFeatures, INDArray trainingS
             //L2
             long stepTime = System.currentTimeMillis();
             if(i % 100 == 0) {
-                LOGGER.info(" grad-step-1:" + (System.currentTimeMillis()-stepTime));
+                LOGGER.info(String.format("Gradient Step 1 computed in %d ms", (System.currentTimeMillis()-stepTime)));
                 stepTime = System.currentTimeMillis();
             }
             double l2 = 1.0 / Transforms.exp(dataPerQueryPredicted.get(keyGen(q, predictedScores)))
                     .sumNumber().doubleValue();
             if(i % 100 == 0) {
-                LOGGER.info(" grad-step-2:" + (System.currentTimeMillis()-stepTime));
+                LOGGER.info(String.format("Gradient Step 2 computed in %d ms", (System.currentTimeMillis()-stepTime)));
                 stepTime = System.currentTimeMillis();
             }
 
@@ -175,7 +175,7 @@ private INDArray calculateGradient(INDArray trainingFeatures, INDArray trainingS
                     .mmul(Transforms.exp(dataPerQueryPredicted.get(keyGen(q, predictedScores))))
                     .mul(l2);
             if(i % 100 == 0) {
-                LOGGER.info(" grad-step-3:" + (System.currentTimeMillis()-stepTime));
+                LOGGER.info(String.format("Gradient Step 3 computed in %d ms", (System.currentTimeMillis()-stepTime)));
                 stepTime = System.currentTimeMillis();
             }
             //L1
@@ -187,7 +187,7 @@ private INDArray calculateGradient(INDArray trainingFeatures, INDArray trainingS
             //L deriv
             res = res.div(Math.log(predictedScores.length()));
             if(i % 100 == 0) {
-                LOGGER.info(" grad-step-4:" + (System.currentTimeMillis()-stepTime));
+                LOGGER.info(String.format("Gradient Step 4 computed in %d ms", (System.currentTimeMillis()-stepTime)));
                 stepTime = System.currentTimeMillis();
             }
             if(!this.noExposure) {
@@ -197,8 +197,7 @@ private INDArray calculateGradient(INDArray trainingFeatures, INDArray trainingS
                         .mul(exposureDiff(predictedScores, queryIds, q, protectedIdxs)).transpose());
             }
             if(i % 100 == 0) {
-                LOGGER.info(" grad-step-5:" + (System.currentTimeMillis()-stepTime));
-                stepTime = System.currentTimeMillis();
+                LOGGER.info(String.format("Gradient Step 5 computed in %d ms", (System.currentTimeMillis()-stepTime)));
             }
             gradient.putRow(i, res);
         });
@@ -225,7 +224,6 @@ private INDArray normalizedToppProtDerivPerGroupDiff(INDArray trainingScores, IN
                     predictionsGroup.getProtectedItemsPerQuery(),
                     predictionsGroup.getJudgementsPerQuery());
 
-//            return u2.sub(u3);
             this.normalizedToppProtDerivPerGroupDiffCache.put(key, u2.sub(u3));
         }
 //