You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@spark.apache.org by Matt Forbes <mf...@twitter.com.INVALID> on 2015/08/13 23:04:44 UTC
Input size increasing every iteration of gradient boosted trees [1.4]
I am training a boosted trees model on a couple million input samples (with
around 300 features) and am noticing that the input size of each stage is
increasing each iteration. For each new tree, the first step seems to be
building the decision tree metadata, which does a .count() on the input
data, so this is the step I've been using to track the input size changing.
Here is what I'm seeing:
count at DecisionTreeMetadata.scala:111
1. Input Size / Records: 726.1 MB / 1295620
2. Input Size / Records: 106.9 GB / 64780816
3. Input Size / Records: 160.3 GB / 97171224
4. Input Size / Records: 214.8 GB / 129680959
5. Input Size / Records: 268.5 GB / 162533424
....
Input Size / Records: 1912.6 GB / 1382017686
....
This step goes from taking less than 10s up to 5 minutes by the 15th or so
iteration. I'm not quite sure what could be causing this. I am passing a
memory-only cached RDD[LabeledPoint] to GradientBoostedTrees.train
Does anybody have some insight? Is this a bug or could it be an error on my
part?
Re: Input size increasing every iteration of gradient boosted trees [1.4]
Posted by Sean Owen <so...@cloudera.com>.
Since it sounds like this has been encountered 3 times, and I've
personally seen it and mostly verified it, I think it's legit enough
for a JIRA: SPARK-10433 I am sorry to say I don't know what is going
here though.
On Thu, Sep 3, 2015 at 1:56 PM, Peter Rudenko <pe...@gmail.com> wrote:
> Confirm, having the same issue (1.4.1 mllib package). For smaller dataset
> accuracy degradeted also. Haven’t tested yet in 1.5 with ml package
> implementation.
>
> val boostingStrategy = BoostingStrategy.defaultParams("Classification")
> boostingStrategy.setNumIterations(30)
> boostingStrategy.setLearningRate(1.0)
> boostingStrategy.treeStrategy.setMaxDepth(3)
> boostingStrategy.treeStrategy.setMaxBins(128)
> boostingStrategy.treeStrategy.setSubsamplingRate(1.0)
> boostingStrategy.treeStrategy.setMinInstancesPerNode(1)
> boostingStrategy.treeStrategy.setUseNodeIdCache(true)
> boostingStrategy.treeStrategy.setCategoricalFeaturesInfo(
>
> mapAsJavaMap(categoricalFeatures).asInstanceOf[java.util.Map[java.lang.Integer,
> java.lang.Integer]])
>
> val model = GradientBoostedTrees.train(instances, boostingStrategy)
>
> Thanks,
> Peter Rudenko
>
> On 2015-08-14 00:33, Sean Owen wrote:
>
> Not that I have any answer at this point, but I was discussing this
> exact same problem with Johannes today. An input size of ~20K records
> was growing each iteration by ~15M records. I could not see why on a
> first look.
>
> @jkbradley I know it's not much info but does that ring any bells? I
> think Johannes even has an instance of this up and running for
> examination.
>
> On Thu, Aug 13, 2015 at 10:04 PM, Matt Forbes
> <mf...@twitter.com.invalid> wrote:
>
> I am training a boosted trees model on a couple million input samples (with
> around 300 features) and am noticing that the input size of each stage is
> increasing each iteration. For each new tree, the first step seems to be
> building the decision tree metadata, which does a .count() on the input
> data, so this is the step I've been using to track the input size changing.
> Here is what I'm seeing:
>
> count at DecisionTreeMetadata.scala:111
> 1. Input Size / Records: 726.1 MB / 1295620
> 2. Input Size / Records: 106.9 GB / 64780816
> 3. Input Size / Records: 160.3 GB / 97171224
> 4. Input Size / Records: 214.8 GB / 129680959
> 5. Input Size / Records: 268.5 GB / 162533424
> ....
> Input Size / Records: 1912.6 GB / 1382017686
> ....
>
> This step goes from taking less than 10s up to 5 minutes by the 15th or so
> iteration. I'm not quite sure what could be causing this. I am passing a
> memory-only cached RDD[LabeledPoint] to GradientBoostedTrees.train
>
> Does anybody have some insight? Is this a bug or could it be an error on my
> part?
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: user-unsubscribe@spark.apache.org
> For additional commands, e-mail: user-help@spark.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@spark.apache.org
For additional commands, e-mail: user-help@spark.apache.org
Re: Input size increasing every iteration of gradient boosted trees [1.4]
Posted by Peter Rudenko <pe...@gmail.com>.
Confirm, having the same issue (1.4.1 mllib package). For smaller
dataset accuracy degradeted also. Haven’t tested yet in 1.5 with ml
package implementation.
|val boostingStrategy = BoostingStrategy.defaultParams("Classification")
boostingStrategy.setNumIterations(30)
boostingStrategy.setLearningRate(1.0)
boostingStrategy.treeStrategy.setMaxDepth(3)
boostingStrategy.treeStrategy.setMaxBins(128)
boostingStrategy.treeStrategy.setSubsamplingRate(1.0)
boostingStrategy.treeStrategy.setMinInstancesPerNode(1)
boostingStrategy.treeStrategy.setUseNodeIdCache(true)
boostingStrategy.treeStrategy.setCategoricalFeaturesInfo(
mapAsJavaMap(categoricalFeatures).asInstanceOf[java.util.Map[java.lang.Integer,
java.lang.Integer]]) val model = GradientBoostedTrees.train(instances,
boostingStrategy) |
Thanks,
Peter Rudenko
On 2015-08-14 00:33, Sean Owen wrote:
> Not that I have any answer at this point, but I was discussing this
> exact same problem with Johannes today. An input size of ~20K records
> was growing each iteration by ~15M records. I could not see why on a
> first look.
>
> @jkbradley I know it's not much info but does that ring any bells? I
> think Johannes even has an instance of this up and running for
> examination.
>
> On Thu, Aug 13, 2015 at 10:04 PM, Matt Forbes
> <mf...@twitter.com.invalid> wrote:
>> I am training a boosted trees model on a couple million input samples (with
>> around 300 features) and am noticing that the input size of each stage is
>> increasing each iteration. For each new tree, the first step seems to be
>> building the decision tree metadata, which does a .count() on the input
>> data, so this is the step I've been using to track the input size changing.
>> Here is what I'm seeing:
>>
>> count at DecisionTreeMetadata.scala:111
>> 1. Input Size / Records: 726.1 MB / 1295620
>> 2. Input Size / Records: 106.9 GB / 64780816
>> 3. Input Size / Records: 160.3 GB / 97171224
>> 4. Input Size / Records: 214.8 GB / 129680959
>> 5. Input Size / Records: 268.5 GB / 162533424
>> ....
>> Input Size / Records: 1912.6 GB / 1382017686
>> ....
>>
>> This step goes from taking less than 10s up to 5 minutes by the 15th or so
>> iteration. I'm not quite sure what could be causing this. I am passing a
>> memory-only cached RDD[LabeledPoint] to GradientBoostedTrees.train
>>
>> Does anybody have some insight? Is this a bug or could it be an error on my
>> part?
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: user-unsubscribe@spark.apache.org
> For additional commands, e-mail: user-help@spark.apache.org
>
Re: Input size increasing every iteration of gradient boosted trees [1.4]
Posted by Matt Forbes <mf...@twitter.com.INVALID>.
Is this an artifact of a recent change? Does this not show up in any of the
tests or benchmarks?
On Thu, Aug 13, 2015 at 2:33 PM, Sean Owen <so...@cloudera.com> wrote:
> Not that I have any answer at this point, but I was discussing this
> exact same problem with Johannes today. An input size of ~20K records
> was growing each iteration by ~15M records. I could not see why on a
> first look.
>
> @jkbradley I know it's not much info but does that ring any bells? I
> think Johannes even has an instance of this up and running for
> examination.
>
> On Thu, Aug 13, 2015 at 10:04 PM, Matt Forbes
> <mf...@twitter.com.invalid> wrote:
> > I am training a boosted trees model on a couple million input samples
> (with
> > around 300 features) and am noticing that the input size of each stage is
> > increasing each iteration. For each new tree, the first step seems to be
> > building the decision tree metadata, which does a .count() on the input
> > data, so this is the step I've been using to track the input size
> changing.
> > Here is what I'm seeing:
> >
> > count at DecisionTreeMetadata.scala:111
> > 1. Input Size / Records: 726.1 MB / 1295620
> > 2. Input Size / Records: 106.9 GB / 64780816
> > 3. Input Size / Records: 160.3 GB / 97171224
> > 4. Input Size / Records: 214.8 GB / 129680959
> > 5. Input Size / Records: 268.5 GB / 162533424
> > ....
> > Input Size / Records: 1912.6 GB / 1382017686
> > ....
> >
> > This step goes from taking less than 10s up to 5 minutes by the 15th or
> so
> > iteration. I'm not quite sure what could be causing this. I am passing a
> > memory-only cached RDD[LabeledPoint] to GradientBoostedTrees.train
> >
> > Does anybody have some insight? Is this a bug or could it be an error on
> my
> > part?
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: user-unsubscribe@spark.apache.org
> For additional commands, e-mail: user-help@spark.apache.org
>
>
Re: Input size increasing every iteration of gradient boosted trees [1.4]
Posted by Sean Owen <so...@cloudera.com>.
Not that I have any answer at this point, but I was discussing this
exact same problem with Johannes today. An input size of ~20K records
was growing each iteration by ~15M records. I could not see why on a
first look.
@jkbradley I know it's not much info but does that ring any bells? I
think Johannes even has an instance of this up and running for
examination.
On Thu, Aug 13, 2015 at 10:04 PM, Matt Forbes
<mf...@twitter.com.invalid> wrote:
> I am training a boosted trees model on a couple million input samples (with
> around 300 features) and am noticing that the input size of each stage is
> increasing each iteration. For each new tree, the first step seems to be
> building the decision tree metadata, which does a .count() on the input
> data, so this is the step I've been using to track the input size changing.
> Here is what I'm seeing:
>
> count at DecisionTreeMetadata.scala:111
> 1. Input Size / Records: 726.1 MB / 1295620
> 2. Input Size / Records: 106.9 GB / 64780816
> 3. Input Size / Records: 160.3 GB / 97171224
> 4. Input Size / Records: 214.8 GB / 129680959
> 5. Input Size / Records: 268.5 GB / 162533424
> ....
> Input Size / Records: 1912.6 GB / 1382017686
> ....
>
> This step goes from taking less than 10s up to 5 minutes by the 15th or so
> iteration. I'm not quite sure what could be causing this. I am passing a
> memory-only cached RDD[LabeledPoint] to GradientBoostedTrees.train
>
> Does anybody have some insight? Is this a bug or could it be an error on my
> part?
---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@spark.apache.org
For additional commands, e-mail: user-help@spark.apache.org