add tokenizer
Browse files- vocab.json +17 -18
vocab.json
CHANGED
|
@@ -1,48 +1,47 @@
|
|
| 1 |
{
|
| 2 |
-
"[PAD]":
|
| 3 |
-
"[UNK]":
|
| 4 |
"|": 27,
|
| 5 |
-
"ĉ": 28,
|
| 6 |
"̣": 8,
|
| 7 |
-
"ء":
|
| 8 |
"آ": 18,
|
| 9 |
"أ": 5,
|
| 10 |
-
"ؤ":
|
| 11 |
"إ": 0,
|
| 12 |
"ئ": 4,
|
| 13 |
"ا": 3,
|
| 14 |
"ب": 2,
|
| 15 |
-
"ة":
|
| 16 |
"ت": 1,
|
| 17 |
"ث": 12,
|
| 18 |
"ج": 9,
|
| 19 |
-
"ح":
|
| 20 |
"خ": 26,
|
| 21 |
"د": 25,
|
| 22 |
"ذ": 20,
|
| 23 |
"ر": 13,
|
| 24 |
"ز": 24,
|
| 25 |
-
"س":
|
| 26 |
-
"ش":
|
| 27 |
"ص": 19,
|
| 28 |
"ض": 17,
|
| 29 |
-
"ط":
|
| 30 |
-
"ظ":
|
| 31 |
"ع": 15,
|
| 32 |
"غ": 23,
|
| 33 |
"ف": 10,
|
| 34 |
"ق": 16,
|
| 35 |
-
"ك":
|
| 36 |
"ل": 22,
|
| 37 |
"م": 11,
|
| 38 |
"ن": 6,
|
| 39 |
-
"ه":
|
| 40 |
-
"و":
|
| 41 |
"ى": 21,
|
| 42 |
-
"ي":
|
| 43 |
-
"ٱ":
|
| 44 |
"چ": 14,
|
| 45 |
-
"ڤ":
|
| 46 |
"ک": 7,
|
| 47 |
-
"ی":
|
| 48 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"[PAD]": 44,
|
| 3 |
+
"[UNK]": 43,
|
| 4 |
"|": 27,
|
|
|
|
| 5 |
"̣": 8,
|
| 6 |
+
"ء": 36,
|
| 7 |
"آ": 18,
|
| 8 |
"أ": 5,
|
| 9 |
+
"ؤ": 33,
|
| 10 |
"إ": 0,
|
| 11 |
"ئ": 4,
|
| 12 |
"ا": 3,
|
| 13 |
"ب": 2,
|
| 14 |
+
"ة": 32,
|
| 15 |
"ت": 1,
|
| 16 |
"ث": 12,
|
| 17 |
"ج": 9,
|
| 18 |
+
"ح": 31,
|
| 19 |
"خ": 26,
|
| 20 |
"د": 25,
|
| 21 |
"ذ": 20,
|
| 22 |
"ر": 13,
|
| 23 |
"ز": 24,
|
| 24 |
+
"س": 29,
|
| 25 |
+
"ش": 37,
|
| 26 |
"ص": 19,
|
| 27 |
"ض": 17,
|
| 28 |
+
"ط": 40,
|
| 29 |
+
"ظ": 28,
|
| 30 |
"ع": 15,
|
| 31 |
"غ": 23,
|
| 32 |
"ف": 10,
|
| 33 |
"ق": 16,
|
| 34 |
+
"ك": 30,
|
| 35 |
"ل": 22,
|
| 36 |
"م": 11,
|
| 37 |
"ن": 6,
|
| 38 |
+
"ه": 38,
|
| 39 |
+
"و": 34,
|
| 40 |
"ى": 21,
|
| 41 |
+
"ي": 35,
|
| 42 |
+
"ٱ": 39,
|
| 43 |
"چ": 14,
|
| 44 |
+
"ڤ": 42,
|
| 45 |
"ک": 7,
|
| 46 |
+
"ی": 41
|
| 47 |
}
|