Upload tokenizer_config.json with huggingface_hub
Browse files- tokenizer_config.json +37 -0
tokenizer_config.json
ADDED
|
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"version": "3.0.0",
|
| 3 |
+
"vocab_size": 32773,
|
| 4 |
+
"pre_tokenizer": "Whitespace",
|
| 5 |
+
"normalizer": "NFKC (NO lowercase)",
|
| 6 |
+
"model": "BPE",
|
| 7 |
+
"brand_tokens": [
|
| 8 |
+
"AksaraLLM",
|
| 9 |
+
"aksarallm",
|
| 10 |
+
"AKSARALLM",
|
| 11 |
+
"Indonesia",
|
| 12 |
+
"indonesia",
|
| 13 |
+
"INDONESIA",
|
| 14 |
+
"Pancasila",
|
| 15 |
+
"pancasila",
|
| 16 |
+
"Nusantara",
|
| 17 |
+
"nusantara"
|
| 18 |
+
],
|
| 19 |
+
"special_tokens": {
|
| 20 |
+
"[PAD]": 0,
|
| 21 |
+
"[EOS]": 1,
|
| 22 |
+
"[BOS]": 2,
|
| 23 |
+
"[UNK]": 3,
|
| 24 |
+
"[SEP]": 4,
|
| 25 |
+
"[MASK]": 5,
|
| 26 |
+
"[INST]": 6,
|
| 27 |
+
"[/INST]": 7,
|
| 28 |
+
"[SYS]": 8,
|
| 29 |
+
"[USER]": 9,
|
| 30 |
+
"[ASST]": 10,
|
| 31 |
+
"[TURN]": 11,
|
| 32 |
+
"[LANG_ID]": 12,
|
| 33 |
+
"[LANG_JV]": 13,
|
| 34 |
+
"[LANG_SU]": 14,
|
| 35 |
+
"[LANG_EN]": 15
|
| 36 |
+
}
|
| 37 |
+
}
|