Repository navigation
Return different matrix types for online serving #4714
Description
Is your feature request related to a problem? Please describe.
We should allow Feature Views to return matrices/tensors natively. For example, torch.tensors.
At the moment, for some features we require the client to serialize the output into a matrix before running inference. Feast should support executing these transformations and serializing the data into matrices for both online and offline retrieval.
Describe the solution you'd like
features: torch.Tensor = store.get_online_features()Describe alternatives you've considered
Not supporting this is the alternative, which is the current state, which leaves users to write their own brittle logic to handle various complexities.
Additional context
@HaoXuAI @tokoko I know we discussed sklearn pipelines in the past and I thought I'd share my thoughts.
Activity
haoxu0 commentedon Oct 29, 2024
torch feature is nice. I guess we need to release the "timestamp" constraints in our APIs, since it probably doesn't make too much sense to attach embedding feature with a timestamp?
breno-costa commentedon Oct 29, 2024
The method
store.get_online_features(...)returns anOnlineResponseobject that has some conversion methods liketo_dict()andto_df(). Should this suggestion be implemented as an another conversion method liketo_torch()or something like this?franciscojavierarceo commentedon Oct 29, 2024
franciscojavierarceo commentedon Oct 29, 2024
@breno-costa that code is a serialization step though. We would want to treat Torch Tensors (or xgb.DMatrix) as a first class data type.
The concrete examples I'm thinking of are one hot encoding or impact encoding. It'd be useful for us to handle this for MLEs natively, especially when handling unseen categories.
dandawg commentedon Oct 29, 2024
This plus sparse tensors/sparse matrices could be a really cool optimization -- less data, faster io, more powerful API.
franciscojavierarceo commentedon Oct 29, 2024
Exactly.
haoxu0 commentedon Oct 29, 2024
if we can leverage "arrow" as our primary format, then it can be directly converted to pandas/torch with arrow apis i believe
franciscojavierarceo commentedon Oct 30, 2024
Cool, I'll check that out. This is basically the next step after vector support to making NLP a first class citizen.
stale commentedon Apr 26, 2025
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.