{ "selected_vocab": 32000, "actual_vocab": 32000, "results": [ { "requested_vocab": 32000, "actual_vocab": 32000, "tokens_per_character": 0.2483463839141529, "tokens_per_whitespace_token": 1.1364876617266, "average_tokens_per_document": 637.061, "probe_fragmentation": { "names": [ { "text": "Nguyễn Ái Quốc", "pieces": [ "▁Nguyễn", "▁Ái", "▁Quốc" ] }, { "text": "Thành phố Hồ Chí Minh", "pieces": [ "▁Thành", "▁phố", "▁Hồ", "▁Chí", "▁Minh" ] }, { "text": "Đắk Lắk", "pieces": [ "▁Đắk", "▁Lắk" ] }, { "text": "OpenAI", "pieces": [ "▁OpenAI" ] } ], "numbers": [ { "text": "1.234.567 đồng", "pieces": [ "▁1.23", "4", ".5", "67", "▁đồng" ] }, { "text": "29/09/2026", "pieces": [ "▁29", "/09", "/2026" ] }, { "text": "3,14159", "pieces": [ "▁3,", "14", "159" ] }, { "text": "2026", "pieces": [ "▁2026" ] } ], "vietnamese": [ { "text": "học sinh", "pieces": [ "▁học", "▁sinh" ] }, { "text": "công nghệ thông tin", "pieces": [ "▁công", "▁nghệ", "▁thông", "▁tin" ] }, { "text": "đất nước", "pieces": [ "▁đất", "▁nước" ] }, { "text": "chuẩn hóa dữ liệu", "pieces": [ "▁chuẩn", "▁hóa", "▁dữ", "▁liệu" ] } ], "english": [ { "text": "machine learning", "pieces": [ "▁ma", "ch", "ine", "▁l", "ear", "ning" ] }, { "text": "artificial intelligence", "pieces": [ "▁art", "ific", "ial", "▁in", "t", "ellig", "ence" ] }, { "text": "New York", "pieces": [ "▁New", "▁York" ] } ] } }, { "requested_vocab": 48000, "actual_vocab": 48000, "tokens_per_character": 0.24706559283708185, "tokens_per_whitespace_token": 1.13062648012452, "average_tokens_per_document": 633.7755, "probe_fragmentation": { "names": [ { "text": "Nguyễn Ái Quốc", "pieces": [ "▁Nguyễn", "▁Ái", "▁Quốc" ] }, { "text": "Thành phố Hồ Chí Minh", "pieces": [ "▁Thành", "▁phố", "▁Hồ", "▁Chí", "▁Minh" ] }, { "text": "Đắk Lắk", "pieces": [ "▁Đắk", "▁Lắk" ] }, { "text": "OpenAI", "pieces": [ "▁OpenAI" ] } ], "numbers": [ { "text": "1.234.567 đồng", "pieces": [ "▁1.23", "4", ".5", "67", "▁đồng" ] }, { "text": "29/09/2026", "pieces": [ "▁29", "/09", "/2026" ] }, { "text": "3,14159", "pieces": [ "▁3,", "14", "159" ] }, { "text": "2026", "pieces": [ "▁2026" ] } ], "vietnamese": [ { "text": "học sinh", "pieces": [ "▁học", "▁sinh" ] }, { "text": "công nghệ thông tin", "pieces": [ "▁công", "▁nghệ", "▁thông", "▁tin" ] }, { "text": "đất nước", "pieces": [ "▁đất", "▁nước" ] }, { "text": "chuẩn hóa dữ liệu", "pieces": [ "▁chuẩn", "▁hóa", "▁dữ", "▁liệu" ] } ], "english": [ { "text": "machine learning", "pieces": [ "▁ma", "ch", "ine", "▁l", "ear", "ning" ] }, { "text": "artificial intelligence", "pieces": [ "▁art", "ificial", "▁in", "t", "ellig", "ence" ] }, { "text": "New York", "pieces": [ "▁New", "▁York" ] } ] } }, { "requested_vocab": 64000, "actual_vocab": 64000, "tokens_per_character": 0.24647148977774347, "tokens_per_whitespace_token": 1.1279077338875485, "average_tokens_per_document": 632.2515, "probe_fragmentation": { "names": [ { "text": "Nguyễn Ái Quốc", "pieces": [ "▁Nguyễn", "▁Ái", "▁Quốc" ] }, { "text": "Thành phố Hồ Chí Minh", "pieces": [ "▁Thành", "▁phố", "▁Hồ", "▁Chí", "▁Minh" ] }, { "text": "Đắk Lắk", "pieces": [ "▁Đắk", "▁Lắk" ] }, { "text": "OpenAI", "pieces": [ "▁OpenAI" ] } ], "numbers": [ { "text": "1.234.567 đồng", "pieces": [ "▁1.23", "4", ".5", "67", "▁đồng" ] }, { "text": "29/09/2026", "pieces": [ "▁29", "/09", "/2026" ] }, { "text": "3,14159", "pieces": [ "▁3,", "14", "159" ] }, { "text": "2026", "pieces": [ "▁2026" ] } ], "vietnamese": [ { "text": "học sinh", "pieces": [ "▁học", "▁sinh" ] }, { "text": "công nghệ thông tin", "pieces": [ "▁công", "▁nghệ", "▁thông", "▁tin" ] }, { "text": "đất nước", "pieces": [ "▁đất", "▁nước" ] }, { "text": "chuẩn hóa dữ liệu", "pieces": [ "▁chuẩn", "▁hóa", "▁dữ", "▁liệu" ] } ], "english": [ { "text": "machine learning", "pieces": [ "▁mach", "ine", "▁l", "ear", "ning" ] }, { "text": "artificial intelligence", "pieces": [ "▁art", "ificial", "▁in", "t", "ellig", "ence" ] }, { "text": "New York", "pieces": [ "▁New", "▁York" ] } ] } } ], "selection_rule": "Smallest vocabulary within 3% of the lowest tokens/character on held-out train documents.", "model": "tokenizer/tokenizer.model", "sha256": "863945d0e4b656eee448fe84fc622e416677a378c6ab88e2153edd7611ab0355", "eos_id": 2, "pad_id": 3, "unk_id": 0, "bos_id": 1, "normalization": "identity; NFC performed upstream; whitespace and paragraph newlines preserved; byte fallback enabled" }