Upload exp_c_tokenizer_ablation.json with huggingface_hub
Browse files- exp_c_tokenizer_ablation.json +203 -0
exp_c_tokenizer_ablation.json
ADDED
|
@@ -0,0 +1,203 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"experiment": "C_tokenizer_ablation",
|
| 3 |
+
"timestamp": "2026-04-13 09:30:52 UTC",
|
| 4 |
+
"tokenizers": {
|
| 5 |
+
"Ours-32K": {
|
| 6 |
+
"vocab_size": 32000
|
| 7 |
+
},
|
| 8 |
+
"Llama-2": {
|
| 9 |
+
"vocab_size": 32000
|
| 10 |
+
},
|
| 11 |
+
"HebrewGPT": {
|
| 12 |
+
"vocab_size": 32000
|
| 13 |
+
}
|
| 14 |
+
},
|
| 15 |
+
"vocabulary_composition": {
|
| 16 |
+
"Ours-32K": {
|
| 17 |
+
"Arabic": {
|
| 18 |
+
"count": 14945,
|
| 19 |
+
"pct": 46.7
|
| 20 |
+
},
|
| 21 |
+
"Hebrew": {
|
| 22 |
+
"count": 8888,
|
| 23 |
+
"pct": 27.8
|
| 24 |
+
},
|
| 25 |
+
"Latin": {
|
| 26 |
+
"count": 7778,
|
| 27 |
+
"pct": 24.3
|
| 28 |
+
},
|
| 29 |
+
"Other": {
|
| 30 |
+
"count": 278,
|
| 31 |
+
"pct": 0.9
|
| 32 |
+
},
|
| 33 |
+
"Digit": {
|
| 34 |
+
"count": 110,
|
| 35 |
+
"pct": 0.3
|
| 36 |
+
},
|
| 37 |
+
"Special": {
|
| 38 |
+
"count": 1,
|
| 39 |
+
"pct": 0.0
|
| 40 |
+
}
|
| 41 |
+
},
|
| 42 |
+
"Llama-2": {
|
| 43 |
+
"Latin": {
|
| 44 |
+
"count": 25900,
|
| 45 |
+
"pct": 80.9
|
| 46 |
+
},
|
| 47 |
+
"Other": {
|
| 48 |
+
"count": 5848,
|
| 49 |
+
"pct": 18.3
|
| 50 |
+
},
|
| 51 |
+
"Digit": {
|
| 52 |
+
"count": 133,
|
| 53 |
+
"pct": 0.4
|
| 54 |
+
},
|
| 55 |
+
"Arabic": {
|
| 56 |
+
"count": 54,
|
| 57 |
+
"pct": 0.2
|
| 58 |
+
},
|
| 59 |
+
"Hebrew": {
|
| 60 |
+
"count": 36,
|
| 61 |
+
"pct": 0.1
|
| 62 |
+
},
|
| 63 |
+
"Special": {
|
| 64 |
+
"count": 16,
|
| 65 |
+
"pct": 0.1
|
| 66 |
+
},
|
| 67 |
+
"Space": {
|
| 68 |
+
"count": 13,
|
| 69 |
+
"pct": 0.0
|
| 70 |
+
}
|
| 71 |
+
},
|
| 72 |
+
"HebrewGPT": {
|
| 73 |
+
"Hebrew": {
|
| 74 |
+
"count": 23101,
|
| 75 |
+
"pct": 72.2
|
| 76 |
+
},
|
| 77 |
+
"Other": {
|
| 78 |
+
"count": 6399,
|
| 79 |
+
"pct": 20.0
|
| 80 |
+
},
|
| 81 |
+
"Latin": {
|
| 82 |
+
"count": 2238,
|
| 83 |
+
"pct": 7.0
|
| 84 |
+
},
|
| 85 |
+
"Arabic": {
|
| 86 |
+
"count": 137,
|
| 87 |
+
"pct": 0.4
|
| 88 |
+
},
|
| 89 |
+
"Digit": {
|
| 90 |
+
"count": 124,
|
| 91 |
+
"pct": 0.4
|
| 92 |
+
},
|
| 93 |
+
"Special": {
|
| 94 |
+
"count": 1,
|
| 95 |
+
"pct": 0.0
|
| 96 |
+
}
|
| 97 |
+
}
|
| 98 |
+
},
|
| 99 |
+
"fertility": {
|
| 100 |
+
"en": {
|
| 101 |
+
"Ours-32K": {
|
| 102 |
+
"fertility": 1.544,
|
| 103 |
+
"bytes_per_token": 3.71,
|
| 104 |
+
"total_tokens": 2785,
|
| 105 |
+
"total_bytes": 10322
|
| 106 |
+
},
|
| 107 |
+
"Llama-2": {
|
| 108 |
+
"fertility": 1.51,
|
| 109 |
+
"bytes_per_token": 3.79,
|
| 110 |
+
"total_tokens": 2724,
|
| 111 |
+
"total_bytes": 10322
|
| 112 |
+
},
|
| 113 |
+
"HebrewGPT": {
|
| 114 |
+
"fertility": 2.419,
|
| 115 |
+
"bytes_per_token": 2.37,
|
| 116 |
+
"total_tokens": 4364,
|
| 117 |
+
"total_bytes": 10322
|
| 118 |
+
}
|
| 119 |
+
},
|
| 120 |
+
"he": {
|
| 121 |
+
"Ours-32K": {
|
| 122 |
+
"fertility": 1.343,
|
| 123 |
+
"bytes_per_token": 5.12,
|
| 124 |
+
"total_tokens": 8866,
|
| 125 |
+
"total_bytes": 45378
|
| 126 |
+
},
|
| 127 |
+
"Llama-2": {
|
| 128 |
+
"fertility": 3.909,
|
| 129 |
+
"bytes_per_token": 1.76,
|
| 130 |
+
"total_tokens": 25806,
|
| 131 |
+
"total_bytes": 45378
|
| 132 |
+
},
|
| 133 |
+
"HebrewGPT": {
|
| 134 |
+
"fertility": 1.255,
|
| 135 |
+
"bytes_per_token": 5.48,
|
| 136 |
+
"total_tokens": 8283,
|
| 137 |
+
"total_bytes": 45378
|
| 138 |
+
}
|
| 139 |
+
},
|
| 140 |
+
"ar": {
|
| 141 |
+
"Ours-32K": {
|
| 142 |
+
"fertility": 2.222,
|
| 143 |
+
"bytes_per_token": 3.48,
|
| 144 |
+
"total_tokens": 7776,
|
| 145 |
+
"total_bytes": 27023
|
| 146 |
+
},
|
| 147 |
+
"Llama-2": {
|
| 148 |
+
"fertility": 4.363,
|
| 149 |
+
"bytes_per_token": 1.77,
|
| 150 |
+
"total_tokens": 15266,
|
| 151 |
+
"total_bytes": 27023
|
| 152 |
+
},
|
| 153 |
+
"HebrewGPT": {
|
| 154 |
+
"fertility": 4.154,
|
| 155 |
+
"bytes_per_token": 1.86,
|
| 156 |
+
"total_tokens": 14535,
|
| 157 |
+
"total_bytes": 27023
|
| 158 |
+
}
|
| 159 |
+
},
|
| 160 |
+
"fa": {
|
| 161 |
+
"Ours-32K": {
|
| 162 |
+
"fertility": 1.52,
|
| 163 |
+
"bytes_per_token": 5.72,
|
| 164 |
+
"total_tokens": 5302,
|
| 165 |
+
"total_bytes": 30327
|
| 166 |
+
},
|
| 167 |
+
"Llama-2": {
|
| 168 |
+
"fertility": 4.876,
|
| 169 |
+
"bytes_per_token": 1.78,
|
| 170 |
+
"total_tokens": 17014,
|
| 171 |
+
"total_bytes": 30327
|
| 172 |
+
},
|
| 173 |
+
"HebrewGPT": {
|
| 174 |
+
"fertility": 4.508,
|
| 175 |
+
"bytes_per_token": 1.93,
|
| 176 |
+
"total_tokens": 15727,
|
| 177 |
+
"total_bytes": 30327
|
| 178 |
+
}
|
| 179 |
+
}
|
| 180 |
+
},
|
| 181 |
+
"efficiency": {
|
| 182 |
+
"en": {
|
| 183 |
+
"Ours-32K": 0.2698,
|
| 184 |
+
"Llama-2": 0.2639,
|
| 185 |
+
"HebrewGPT": 0.4228
|
| 186 |
+
},
|
| 187 |
+
"he": {
|
| 188 |
+
"Ours-32K": 0.1954,
|
| 189 |
+
"Llama-2": 0.5687,
|
| 190 |
+
"HebrewGPT": 0.1825
|
| 191 |
+
},
|
| 192 |
+
"ar": {
|
| 193 |
+
"Ours-32K": 0.2878,
|
| 194 |
+
"Llama-2": 0.5649,
|
| 195 |
+
"HebrewGPT": 0.5379
|
| 196 |
+
},
|
| 197 |
+
"fa": {
|
| 198 |
+
"Ours-32K": 0.1748,
|
| 199 |
+
"Llama-2": 0.561,
|
| 200 |
+
"HebrewGPT": 0.5186
|
| 201 |
+
}
|
| 202 |
+
}
|
| 203 |
+
}
|