Hi alekandreev, Any reason you decided to go with a token vocabulary size of 256k? Smaller vocab/vector sizes like most models in this size seem to be using (~

Hi alekandreev, Any reason you decided to go with a token vocabulary size of 256k? Smaller vocab/vector sizes like most models in this size seem to be using (~16-32k) are much easier to work with. Would love to understand the technical reasoning here that isn't detailed in the report unfortunately :(.