nicola007b

Is there any update on using the new parameter server based implementation of Word2vec? Can anyone share a working example ? Thanks

Thanks for the prompt reply. Is the ‘old implementation’ still the recommended way of running word2vec on spark ? Any advice on this ?

Hi @namitk, is there any update on this issue? I tried to reimplement `TopNQueue` using java.util.PriorityQueue but still got Serialization issues. You can see the gist: https://gist.github.com/nicola007b/454bc77c435cff65e5cdd73ced316e1c ``` 18/10/04 07:48:59...

@oscaroboto thanks for following up on this. I believe the issue is in https://github.com/linkedin/scanns/blob/master/scanns/src/main/scala/com/linkedin/nn/model/LSHNearestNeighborSearchModel.scala#L55 which takes as input an `Iterator[Array[mutable.ArrayBuffer[ItemId]]]` which is not serializable. @namitk what are your thoughts on...

@oscaroboto @Itfly would be great if you could open a PR with the fix