ParentFull threadniyazpk·>> We've seen ~200MB production models get down to ~4MB and not lose much precision.Details please. What techniques are used to reduce the model size?View on HN