Jinjing Zhou

I think the major reason is `SELECT * FROM documents WHERE doc_emb $1 < 0.2` cannot go through the index. This is a known limitation of postgres. Our solution is...

pgvector also had a similar discussion on the distance filter problem at https://github.com/pgvector/pgvector/issues/678

We have patched it in https://github.com/tensorchord/pgvecto.rs/pull/588. However updating version with immich database is quite complicated. We'll work with immich when we have a new production release ready

@beyondwatts Are you experiencing problem at restore or backup?

@hagak You can just drop all the vector index and rebuild them.

Hi, Thanks for your interest. We had some initial test using pgvecto.rs with distributed fdw and it works. I think citus is a similar case. And we're actually developing https://github.com/tensorchord/VectorChord,...

For postgres, you can easily add more replica for QPS requirement. For the distributed idea, I think it's valid and you get the point of ivfflat. But the problem is...

Hi, we have a branch to support PG18 and already tested it with our pgrx extension. Check it at https://github.com/silver-ymz/pgrx/tree/support-18-devel cc @silver-ymz

It's weird, did you see D2H copy without the forward process? Also by setting ` tf.config.experimental.set_device_policy("explicit")`, got error ``` InvalidArgumentError: Tensors on conflicting devices: cannot compute LogicalNot as input #0...

I found the error with `tf.config.experimental.set_device_policy("explicit")` is not the error we are looking for. It's some Keras internal issues. Actually I cannot reproduce the error with ```python import cupy as...