0.0.8

Flops scheduler and max aggregation over language models logins.

0.0.7

Set k_tokens parameters for both Splade (retriever) and SparsEmbed (model and retriever) in order to filter activated tokens.

0.0.6

Optimize retrievers.

0.0.5

Update requirements and tests.

0.0.4

Include splade model and wrap training.

0.0.3

Update scoring function.

0.0.2

Rely on torch sparse tensors for the retriever.

0.0.1

SparsEmbed 0.01 version. 👋