HyukjinKwon commented on a change in pull request #32673:
URL: https://github.com/apache/spark/pull/32673#discussion_r640329780
##########
File path:
mllib/src/test/scala/org/apache/spark/ml/optim/aggregator/DifferentiableLossAggregatorSuite.scala
##########
@@ -149,10 +152,43 @@ object DifferentiableLossAggregatorSuite {
override def add(instance: Instance): TestAggregator = {
val error = instance.label - BLAS.dot(coefficients, instance.features)
- weightSum += instance.weight
- lossSum += instance.weight * error * error / 2.0
- (0 until dim).foreach { j =>
- gradientSumArray(j) += instance.weight * error * instance.features(j)
+ val mathCls = Utils.classForName("java.lang.Math")
Review comment:
@sarutak, I think we can just fix the test with tolerance e.g.)
`assert(agg.weight ~== weightSum relTol 1e-5)` like the above one or
https://github.com/apache/spark/blob/master/mllib/src/test/scala/org/apache/spark/ml/tuning/RandomRangesSuite.scala#L37
Up to my knowledge, this kind of precision diff is a common problem in ML
side. cc @srowen, @WeichenXu123 @zhengruifeng FYI
##########
File path:
mllib/src/test/scala/org/apache/spark/ml/optim/aggregator/DifferentiableLossAggregatorSuite.scala
##########
@@ -149,10 +152,43 @@ object DifferentiableLossAggregatorSuite {
override def add(instance: Instance): TestAggregator = {
val error = instance.label - BLAS.dot(coefficients, instance.features)
- weightSum += instance.weight
- lossSum += instance.weight * error * error / 2.0
- (0 until dim).foreach { j =>
- gradientSumArray(j) += instance.weight * error * instance.features(j)
+ val mathCls = Utils.classForName("java.lang.Math")
Review comment:
@sarutak, I think we can just fix the test with tolerance at the
assertion e.g.) `assert(agg.weight ~== weightSum relTol 1e-5)` like the above
one or
https://github.com/apache/spark/blob/master/mllib/src/test/scala/org/apache/spark/ml/tuning/RandomRangesSuite.scala#L37
Up to my knowledge, this kind of precision diff is a common problem in ML
side. cc @srowen, @WeichenXu123 @zhengruifeng FYI
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
