You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is your feature request related to a problem?
Currently vectors are represented as floats(4 Bytes) for each dimension. This could lead to large storage and memory requirements to construct/load/save/search graphs and at scale becomes costly and also hard to manage the cluster requirements. In some cases, customers prefer to reduce the storage/memory foot prints at the cost of loosing tolerable recall.
What solution would you like?
We could provide Byte sized vectors that can use one Byte per dimension instead of current 4 Bytes using the ByteVector feature of Lucene. By providing quantization techniques we could let the floats convert to Int[-128 to 127] at loss of some precision and hence provide reduction to storage and memory requirements
The text was updated successfully, but these errors were encountered:
Is your feature request related to a problem?
Currently vectors are represented as floats(4 Bytes) for each dimension. This could lead to large storage and memory requirements to construct/load/save/search graphs and at scale becomes costly and also hard to manage the cluster requirements. In some cases, customers prefer to reduce the storage/memory foot prints at the cost of loosing tolerable recall.
What solution would you like?
We could provide Byte sized vectors that can use one Byte per dimension instead of current 4 Bytes using the ByteVector feature of Lucene. By providing quantization techniques we could let the floats convert to Int[-128 to 127] at loss of some precision and hence provide reduction to storage and memory requirements
The text was updated successfully, but these errors were encountered: