Yes, let's make a note of it and proceed till we get some help from H2O.

On Tue, Jul 7, 2015 at 5:13 PM, Thushan Ganegedara <[email protected]> wrote:

> Hello,
>
> It seems it's quite hard to debug the issue with custom serialization of
> DLModel and there's no reply from developers either.
>
> One possibility is to handle the persistence and retrieval for the DLModel
> in the MLModelHandler itself, until this gets sorted out.
>
> Could you share your opinion on this?
>
> On Tue, Jul 7, 2015 at 12:01 PM, Thushan Ganegedara <[email protected]>
> wrote:
>
>> Hi all,
>>
>> I've introduced a MLModel for Deeplearning and currently looking into
>> serialization of that model. It seems serialization of H-2-O
>> DeeplearningModel (with custom serializers) is quite tricky.
>>
>> I have been able to successfully serialize the model by using an
>> AutoBuffer (H-2-O) and ByteStream
>>
>> However in the deserialization process, it gives an error (described in
>> https://groups.google.com/forum/#!topic/h2ostream/JRjdj2f7CXw)
>>
>> Therefore it is still impossible to deserialize the model. I've done
>> extensive debuggin on H-2-O tracking all the steps during the
>> serialization. But it seems to occur from a class created during start up
>> time by something called a "Weaver". I still trying to figure out how to
>> debug them.
>>
>> Thank you
>>
>> --
>> Regards,
>>
>> Thushan Ganegedara
>> School of IT
>> University of Sydney, Australia
>>
>
>
>
> --
> Regards,
>
> Thushan Ganegedara
> School of IT
> University of Sydney, Australia
>



-- 

Thanks & regards,
Nirmal

Associate Technical Lead - Data Technologies Team, WSO2 Inc.
Mobile: +94715779733
Blog: http://nirmalfdo.blogspot.com/
_______________________________________________
Dev mailing list
[email protected]
http://wso2.org/cgi-bin/mailman/listinfo/dev

Reply via email to