Comment by tjgreen

For 2M scale dataset, you should be able to index in about 1 minute on low-end hardware. See the MS-MARCO v1 (8M documents) numbers, measured on cheap Github runners.