guys have anyone try recommendations? I try to gen...
# community-help
d
guys have anyone try recommendations? I try to generate embedding from model that I training, but why the all embedding is 0
k
Please post the full reproducible code.
d
So I train to generate embedding ./starspace train -trainFile /research/Starspace/spes_sub_spes.txt -model SpesialisModel -label '' -trainMode 1 -epoch 50 -dim 768 and then I try to generate embedding the data train while read -r item; do embedding=$(echo "${item}" | ./embed_doc SpesialisModel| tail -1 | tr ' ' ',') echo "{\"id\":\"${item}\",\"embedding\":[${embedding%?}]}" >> "${output_jsonl_file}" done <<< "${unique_items}" then, I try to generate data test or another input export embedding=$(echo "laktasi" | ./embed_doc SpesialisModel | tail -1 | tr ' ' ',') but I got the embedding 0.
So I'm trying to make a recommendation for vector search which is not include in dataset, is it possible to use recommendation
I have some data want to recommend product. So if the user want to search the data in not in the dataset, my expected is can be recommend another product
@Kishore Nallan Can u help me with this?
k
The starspace related guide is a bit outdated now because that project is no longer maintained by Facebook. We are working on a new recommendation engine. Meanwhile I recommend relying on content based recommendations by embedding of the content.