The setting of hyper-parameter "n_samples_inference" is 7 may incorrect? #15
Description
If the hyper-parameter is set as 7, the number of samples in the retrieval pool will be also increased 7 times. When the retrieval sample for R@1 (top1) is predicted wrongly, then the evaluation of R@5 is also predicted wrongly. Is the evaluation method wrong?