Ignite ML withKeepBinary cache

classic Classic list List threaded Threaded
5 messages Options
otorreno otorreno
Reply | Threaded
Open this post in threaded view
|

Ignite ML withKeepBinary cache

Hi everyone,

After the new release (2.7.0), I have been playing around with the machine
learning algorithms a bit.
We have some data in a cache created with the "withKeepBinary()" option, and
I wanted
to test if the machine learning algos would work with such a cache. I tried,
but it fails with the following stacktrace:

org.apache.ignite.IgniteException: testType
    at
org.apache.ignite.internal.processors.closure.GridClosureProcessor$C2.execute(GridClosureProcessor.java:1858)
    at
org.apache.ignite.internal.processors.job.GridJobWorker$2.call(GridJobWorker.java:568)
    at
org.apache.ignite.internal.util.IgniteUtils.wrapThreadLoader(IgniteUtils.java:6816)
    at
org.apache.ignite.internal.processors.job.GridJobWorker.execute0(GridJobWorker.java:562)
    at
org.apache.ignite.internal.processors.job.GridJobWorker.body(GridJobWorker.java:491)
    at
org.apache.ignite.internal.util.worker.GridWorker.run(GridWorker.java:120)
    at
java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
    at
java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
    at java.lang.Thread.run(Thread.java:748)
Caused by: org.apache.ignite.binary.BinaryInvalidTypeException: testType
    at
org.apache.ignite.internal.binary.BinaryContext.descriptorForTypeId(BinaryContext.java:707)
    at
org.apache.ignite.internal.binary.BinaryReaderExImpl.deserialize0(BinaryReaderExImpl.java:1757)
    at
org.apache.ignite.internal.binary.BinaryReaderExImpl.deserialize(BinaryReaderExImpl.java:1716)
    at
org.apache.ignite.internal.binary.BinaryObjectImpl.deserializeValue(BinaryObjectImpl.java:798)
    at
org.apache.ignite.internal.binary.BinaryObjectImpl.value(BinaryObjectImpl.java:143)
    at
org.apache.ignite.internal.processors.cache.CacheObjectUtils.unwrapBinary(CacheObjectUtils.java:177)
    at
org.apache.ignite.internal.processors.cache.CacheObjectUtils.unwrapBinaryIfNeeded(CacheObjectUtils.java:39)
    at
org.apache.ignite.internal.processors.cache.query.GridCacheQueryManager$ScanQueryIterator.advance(GridCacheQueryManager.java:3063)
    at
org.apache.ignite.internal.processors.cache.query.GridCacheQueryManager$ScanQueryIterator.onHasNext(GridCacheQueryManager.java:2965)
    at
org.apache.ignite.internal.util.GridCloseableIteratorAdapter.hasNextX(GridCloseableIteratorAdapter.java:53)
    at
org.apache.ignite.internal.util.lang.GridIteratorAdapter.hasNext(GridIteratorAdapter.java:45)
    at
org.apache.ignite.ml.dataset.impl.cache.util.ComputeUtils.computeCount(ComputeUtils.java:313)
    at
org.apache.ignite.ml.dataset.impl.cache.util.ComputeUtils.computeCount(ComputeUtils.java:300)
    at
org.apache.ignite.ml.dataset.impl.cache.util.ComputeUtils.lambda$initContext$9b68d858$1(ComputeUtils.java:222)
    at
org.apache.ignite.ml.dataset.impl.cache.util.ComputeUtils.lambda$affinityCallWithRetries$b46c4136$1(ComputeUtils.java:90)
    at
org.apache.ignite.internal.processors.closure.GridClosureProcessor$C2.execute(GridClosureProcessor.java:1855)
    ... 8 common frames omitted
Caused by: java.lang.ClassNotFoundException: testType
    at java.net.URLClassLoader.findClass(URLClassLoader.java:381)
    at java.lang.ClassLoader.loadClass(ClassLoader.java:424)
    at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:338)
    at java.lang.ClassLoader.loadClass(ClassLoader.java:357)
    at java.lang.Class.forName0(Native Method)
    at java.lang.Class.forName(Class.java:348)
    at
org.apache.ignite.internal.util.IgniteUtils.forName(IgniteUtils.java:8771)
    at
org.apache.ignite.internal.MarshallerContextImpl.getClass(MarshallerContextImpl.java:349)
    at
org.apache.ignite.internal.binary.BinaryContext.descriptorForTypeId(BinaryContext.java:698)
    ... 23 common frames omitted

Debugging, I found the source of the error, at some point you are just
taking the
name of the upstreamCache (where the data resides), and creating a new
IgniteCache
object using such name before copying the data to a dataset cache. However,
you
are not using the keepBinary property of the original cache. I hardcoded the
"withKeepBinary()" to the following lines:
https://github.com/apache/ignite/blob/2.7.0/modules/ml/src/main/java/org/apache/ignite/ml/dataset/impl/cache/util/ComputeUtils.java#L162
https://github.com/apache/ignite/blob/2.7.0/modules/ml/src/main/java/org/apache/ignite/ml/dataset/impl/cache/util/ComputeUtils.java#L215
https://github.com/apache/ignite/blob/2.7.0/modules/ml/src/main/java/org/apache/ignite/ml/dataset/impl/cache/CacheBasedDatasetBuilder.java#L99

The previous made it work. I tried to retrieve the keep binary property from
the
upstreamCache, but I was not able to find the right method to obtain it (I
saw the property is
stored in the operation context field (opCtx), but it is private and cannot
be
accessed from the lines I modified)

My example code is available at:
https://gist.github.com/otorreno/ca6c5347c1bbde2d4fedd02b51d02cbb

Any plans on making the machine learning algorithms work with caches with
keepBinary set to true?



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/
stephendarlington stephendarlington
Reply | Threaded
Open this post in threaded view
|

Re: Ignite ML withKeepBinary cache

That’s a great investigation! I think the developer mailing list (http://apache-ignite-developers.2346864.n4.nabble.com) would be a better place to discuss the best way to fix it, though.

Regards,
Stephen

> On 2 Jan 2019, at 07:20, otorreno <[hidden email]> wrote:
>
> Hi everyone,
>
> After the new release (2.7.0), I have been playing around with the machine
> learning algorithms a bit.
> We have some data in a cache created with the "withKeepBinary()" option, and
> I wanted
> to test if the machine learning algos would work with such a cache. I tried,
> but it fails with the following stacktrace:
>
> org.apache.ignite.IgniteException: testType
>    at
> org.apache.ignite.internal.processors.closure.GridClosureProcessor$C2.execute(GridClosureProcessor.java:1858)
>    at
> org.apache.ignite.internal.processors.job.GridJobWorker$2.call(GridJobWorker.java:568)
>    at
> org.apache.ignite.internal.util.IgniteUtils.wrapThreadLoader(IgniteUtils.java:6816)
>    at
> org.apache.ignite.internal.processors.job.GridJobWorker.execute0(GridJobWorker.java:562)
>    at
> org.apache.ignite.internal.processors.job.GridJobWorker.body(GridJobWorker.java:491)
>    at
> org.apache.ignite.internal.util.worker.GridWorker.run(GridWorker.java:120)
>    at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
>    at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
>    at java.lang.Thread.run(Thread.java:748)
> Caused by: org.apache.ignite.binary.BinaryInvalidTypeException: testType
>    at
> org.apache.ignite.internal.binary.BinaryContext.descriptorForTypeId(BinaryContext.java:707)
>    at
> org.apache.ignite.internal.binary.BinaryReaderExImpl.deserialize0(BinaryReaderExImpl.java:1757)
>    at
> org.apache.ignite.internal.binary.BinaryReaderExImpl.deserialize(BinaryReaderExImpl.java:1716)
>    at
> org.apache.ignite.internal.binary.BinaryObjectImpl.deserializeValue(BinaryObjectImpl.java:798)
>    at
> org.apache.ignite.internal.binary.BinaryObjectImpl.value(BinaryObjectImpl.java:143)
>    at
> org.apache.ignite.internal.processors.cache.CacheObjectUtils.unwrapBinary(CacheObjectUtils.java:177)
>    at
> org.apache.ignite.internal.processors.cache.CacheObjectUtils.unwrapBinaryIfNeeded(CacheObjectUtils.java:39)
>    at
> org.apache.ignite.internal.processors.cache.query.GridCacheQueryManager$ScanQueryIterator.advance(GridCacheQueryManager.java:3063)
>    at
> org.apache.ignite.internal.processors.cache.query.GridCacheQueryManager$ScanQueryIterator.onHasNext(GridCacheQueryManager.java:2965)
>    at
> org.apache.ignite.internal.util.GridCloseableIteratorAdapter.hasNextX(GridCloseableIteratorAdapter.java:53)
>    at
> org.apache.ignite.internal.util.lang.GridIteratorAdapter.hasNext(GridIteratorAdapter.java:45)
>    at
> org.apache.ignite.ml.dataset.impl.cache.util.ComputeUtils.computeCount(ComputeUtils.java:313)
>    at
> org.apache.ignite.ml.dataset.impl.cache.util.ComputeUtils.computeCount(ComputeUtils.java:300)
>    at
> org.apache.ignite.ml.dataset.impl.cache.util.ComputeUtils.lambda$initContext$9b68d858$1(ComputeUtils.java:222)
>    at
> org.apache.ignite.ml.dataset.impl.cache.util.ComputeUtils.lambda$affinityCallWithRetries$b46c4136$1(ComputeUtils.java:90)
>    at
> org.apache.ignite.internal.processors.closure.GridClosureProcessor$C2.execute(GridClosureProcessor.java:1855)
>    ... 8 common frames omitted
> Caused by: java.lang.ClassNotFoundException: testType
>    at java.net.URLClassLoader.findClass(URLClassLoader.java:381)
>    at java.lang.ClassLoader.loadClass(ClassLoader.java:424)
>    at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:338)
>    at java.lang.ClassLoader.loadClass(ClassLoader.java:357)
>    at java.lang.Class.forName0(Native Method)
>    at java.lang.Class.forName(Class.java:348)
>    at
> org.apache.ignite.internal.util.IgniteUtils.forName(IgniteUtils.java:8771)
>    at
> org.apache.ignite.internal.MarshallerContextImpl.getClass(MarshallerContextImpl.java:349)
>    at
> org.apache.ignite.internal.binary.BinaryContext.descriptorForTypeId(BinaryContext.java:698)
>    ... 23 common frames omitted
>
> Debugging, I found the source of the error, at some point you are just
> taking the
> name of the upstreamCache (where the data resides), and creating a new
> IgniteCache
> object using such name before copying the data to a dataset cache. However,
> you
> are not using the keepBinary property of the original cache. I hardcoded the
> "withKeepBinary()" to the following lines:
> https://github.com/apache/ignite/blob/2.7.0/modules/ml/src/main/java/org/apache/ignite/ml/dataset/impl/cache/util/ComputeUtils.java#L162
> https://github.com/apache/ignite/blob/2.7.0/modules/ml/src/main/java/org/apache/ignite/ml/dataset/impl/cache/util/ComputeUtils.java#L215
> https://github.com/apache/ignite/blob/2.7.0/modules/ml/src/main/java/org/apache/ignite/ml/dataset/impl/cache/CacheBasedDatasetBuilder.java#L99
>
> The previous made it work. I tried to retrieve the keep binary property from
> the
> upstreamCache, but I was not able to find the right method to obtain it (I
> saw the property is
> stored in the operation context field (opCtx), but it is private and cannot
> be
> accessed from the lines I modified)
>
> My example code is available at:
> https://gist.github.com/otorreno/ca6c5347c1bbde2d4fedd02b51d02cbb
>
> Any plans on making the machine learning algorithms work with caches with
> keepBinary set to true?
>
>
>
> --
> Sent from: http://apache-ignite-users.70518.x6.nabble.com/


ilya.kasnacheev ilya.kasnacheev
Reply | Threaded
Open this post in threaded view
|

Re: Ignite ML withKeepBinary cache

Hello!

I can see there is a ticket in JIRA:

https://issues.apache.org/jira/browse/IGNITE-10700

If anything is not there please update it.

Regards,



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/
otorreno otorreno
Reply | Threaded
Open this post in threaded view
|

Re: Ignite ML withKeepBinary cache

Thanks Ilya,

I got the link to the JIRA ticket in the Ignite Devs mailing list. In fact I
already included a comment in the ticket.

I got a response from Alexey Zinoviev in the Dev list too, and I am now
waiting to receive a further update from him on this matter.

Regards,



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/
Yury Babak Yury Babak
Reply | Threaded
Open this post in threaded view
|

Re: Ignite ML withKeepBinary cache

Hi all,

Ticket 10700 [1] is resolved, this ticked added support of training models over a cache with binary objects(cache with enabled keepBinary flag) for more details please take a look the mentioned ticked or added example [2].

[2] - org.apache.ignite.examples.ml.TrainingWithBinaryObjectExample

Best regards,
Yuriy Babak


пн, 21 янв. 2019 г. в 11:13, otorreno <[hidden email]>:
Thanks Ilya,

I got the link to the JIRA ticket in the Ignite Devs mailing list. In fact I
already included a comment in the ticket.

I got a response from Alexey Zinoviev in the Dev list too, and I am now
waiting to receive a further update from him on this matter.

Regards,



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/