Maciej Kula
Maciej Kula
Thanks for the comprehensive answer, Patrick!
Thanks for the question! I have some workaround below but I agree that the ergonomics here aren't great. I'll investigate further and see if this can be made better. The...
@cfregly if you'd like to export a brute-force based model, you'll need to pick out the retrieval model subcomponents. ```python # Create a model that takes in raw query features,...
I suspect that you're passing a nested dict of tensors into your function, and it has the `customer_id` key repeated more than once. When saving models, these are flattened, and...
Does this algorithm even make sense when there are no indices? Maybe it would be preferable to return an error.
For BPR we want dot products, not cosine similarity (at least it's defined that way in the paper). `K.batch_dot` may still be useful, though.
You are quite right. Would you be willing to make a PR? Otherwise I'll stick this in my backlog and get round to it when I have time.
Also: thank you for the amazing work on Serde. I always use it as a key selling point in my Rust proselytising.
All the items that the user could have interacted with (bought/rated) are implicitly treated as negatives.
Maybe you added too much regularization?