Training by dot product and evaluation via inverted index? about splade HOT 2 CLOSED

Hey,
Thanks for your interest in SPLADE !

1. The model is indeed trained with dot product
2. We have some internal code which relies on custom inverted index with numba for indexing and retrieval; we should release it soon. But as you said, you can also use Anserini, following the guidelines here. Note that the two arguments -impact -pretokenized tell anserini to use dot product.
3. Yes; we should also refactor this part too
4. Note that the model relies on a ReLU activation; thus, training with L1/FLOPS directly makes the representations sparse (hence, no need for a threshold) => one forward with the model outputs sparse representations

Best,
Thibault

from splade.

That's clear. Thank you for your reply.

from splade.