By the way, I am doing this on Solr 7.4.0. Sorry, forgot to include that in my earlier post.
On Wed, 5 Sep 2018 at 17:39, Zheng Lin Edwin Yeo <edwinye...@gmail.com> wrote: > Hi, > > I am looking at implementing the NeuralNetworkModel for the Solr LTR which > uses the "pairwise" approach, as it seems to be a better option than the > LinearModel, which uses the "pointwise" approach.. > > I am following the setup from the following site: > https://github.com/airalcorn2/Solr-LTR#RankNet to setup RankNet. However, > I am not sure of the codes in the example, like which code should we run > first? > > The first code below seems to be already calling the model, but we have > not train and upload the model yet, and those codes are in the later > section. > > url = "http://localhost:8983/solr/test/query"; > url += "?q={0}&df=text&rq={{!ltr model=my_efi_model ".format(text) > url += "efi.text_a='{0}' efi.text_b='{1}' efi.text='{2}'}}".format(text_a, > text_b, text) > url += "&fl=id,score,[features]&rows={1}".format(text, RERANK) > > > Anyone has tried this before and know of the correct sequence to setup the > RankNet successfully? > > Regards, > Edwin > >
