possible to integrate with vidore bench codebase
#6
by Buluchacha - opened
Hi Team,
Congrats for your great work! Currently we are evaluating all multi-vector retrievers in vidore benchmark (see: https://github.com/illuin-tech/vidore-benchmark), and we surprisingly find your lama-nemoretriever-colembed series are the top models on Vidore-v2. Thus we hope you can implement your retriever in this codebase (see: https://github.com/illuin-tech/vidore-benchmark/tree/main/src/vidore_benchmark/retrievers), as llama-3.2 as llm backbone is never adopted by other retrievers yet. We look forward to your kind reply and contribution!