Hello, Yes, I also think sticking to a previous release might be the best option here, as all the latest releases seems to be unstable in many aspects. For example some DL4J dependencies are still very primitive (e.g. Canova has only 1 release - 0.0.0.1)
On Fri, Mar 27, 2015 at 3:49 PM, Nirmal Fernando <[email protected]> wrote: > Hi Thushan, > > Great work on getting this to work! IMO we should stick to a previous > release till we get a stable new release. We can't keep a locally built one > due to the maintenance overhead. > > On Fri, Mar 27, 2015 at 10:07 AM, Thushan Ganegedara <[email protected]> > wrote: > >> Hello, >> >> I was able to get the DeepLearning4J example working ( >> https://github.com/deeplearning4j/dl4j-0.0.3.3-examples) after some >> extensive building and dependency updates. I haven't still run anything as >> I need to bring them to produce proper outputs, but I was able to compile >> WSO2-ML without any issues while including the deeplearning4j MNIST example. >> >> Just for the record I had to follow the following procedure in order to >> get this working >> >> Download *Canova *from (https://github.com/deeplearning4j/Canova.git) >> and build the project and install dependency via mvn install:install-file >> However, in order to compile successfully, I had to skip the Tests and >> Javadoc check because otherwise I cannot compile it (figured out this after >> a discussion with one of the contributors) >> >> Download *nd4j *from (https://github.com/deeplearning4j/nd4j) and build >> it and install it via mvn install:install-file. Because the any of the >> releases do not have, do not contain the (nd4j.linalg.dataset.api.iterator) >> >> Download *DeepLearning4J *from ( >> https://github.com/deeplearning4j/deeplearning4j) and build it and >> install it via mvn install:install-file. Because the any of the releases do >> not have org.deepleaning4j.nn.conf.override package >> >> After the process I have following versions >> >> - Canova - 0.0.0.2-SNAPSHOT >> - nd4j - 0.0.3.5.5.3-SNAPSHOT >> - dl4j - 0.0.3.3.3.alpha1-SNAPSHOT >> >> >> In these conditions it compiles without any issues. >> >> So I guess we have 2 major options >> >> - continue with the locally built dependencies or >> - switch to a previous release which supports this example ( >> http://deeplearning4j.org/mnist-tutorial.html) >> >> >> Thank you >> >> >> On Fri, Mar 27, 2015 at 7:03 AM, Thushan Ganegedara <[email protected]> >> wrote: >> >>> We can try to maybe use a previous release corresponding to the example >>> at http://deeplearning4j.org/mnist-tutorial.html (may be <= 0.0.3.2) >>> However, as the explain this in there github issue >>> https://github.com/deeplearning4j/deeplearning4j/issues/173 using it >>> reduces the generality of the deepLearning. But I think it can get the work >>> done. >>> >>> >>> >>> On Thu, Mar 26, 2015 at 11:04 PM, Nirmal Fernando <[email protected]> >>> wrote: >>> >>>> Can't we use a released version and samples corresponds to a released >>>> version? >>>> >>>> On Thu, Mar 26, 2015 at 5:29 PM, Thushan Ganegedara <[email protected]> >>>> wrote: >>>> >>>>> Dear Devs, >>>>> >>>>> For the past few days, I've been try to add deeplearning4j dependency >>>>> with maven and was able to do it successfully. >>>>> >>>>> However, the example provided in >>>>> http://deeplearning4j.org/mnist-tutorial.html uses a class called *DBN >>>>> *which is no longer available in the latest deeplearning4j releases. >>>>> >>>>> Furthermore, they have other examples available at >>>>> https://github.com/deeplearning4j/dl4j-0.0.3.3-examples which again >>>>> uses a class (i.e. org.deeplearning4j.nn.conf.override. >>>>> *ClassifierOverride*) which is not available in any of the releases, >>>>> but only the source. I asked about this from the contributors and they >>>>> told >>>>> me to build from the source. >>>>> >>>>> But when I tried that, I was unable to get it to compile, as Canova >>>>> (0.0.0.2) cannot find one of it's dependencies ( >>>>> *openjfx-78-backport-native*) and it's not mentioned in Canova's >>>>> pom.xml Therefore I tried including it in the pom.xml and compiling it. >>>>> But >>>>> it's still giving me some problems. >>>>> >>>>> So by looking at all these information, deeplearning4j seems a little >>>>> bit unstable at the moment. >>>>> >>>>> I would highly appreciate your thoughts about this and if there are >>>>> any alternative suggestions. >>>>> >>>>> Thank you >>>>> >>>>> >>>>> -- >>>>> Regards, >>>>> >>>>> Thushan Ganegedara >>>>> School of IT >>>>> University of Sydney, Australia >>>>> >>>> >>>> >>>> >>>> -- >>>> >>>> Thanks & regards, >>>> Nirmal >>>> >>>> Senior Software Engineer- Platform Technologies Team, WSO2 Inc. >>>> Mobile: +94715779733 >>>> Blog: http://nirmalfdo.blogspot.com/ >>>> >>>> >>>> >>> >>> >>> -- >>> Regards, >>> >>> Thushan Ganegedara >>> School of IT >>> University of Sydney, Australia >>> >> >> >> >> -- >> Regards, >> >> Thushan Ganegedara >> School of IT >> University of Sydney, Australia >> > > > > -- > > Thanks & regards, > Nirmal > > Senior Software Engineer- Platform Technologies Team, WSO2 Inc. > Mobile: +94715779733 > Blog: http://nirmalfdo.blogspot.com/ > > > -- Regards, Thushan Ganegedara School of IT University of Sydney, Australia
_______________________________________________ Dev mailing list [email protected] http://wso2.org/cgi-bin/mailman/listinfo/dev
