build guide 1. download tokenizer model from hf python ./scripts/download.py 2. export tokenizer model to json format python ./scripts/export_tokenizer_model.py 3. build all vocab code bash ./scripts/build-all-vocab.bash 4. build all packages pnpm build