Upload .\data\processed\collection_summary.json with huggingface_hub
Browse files
.//data//processed//collection_summary.json
ADDED
|
@@ -0,0 +1,284 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"huggingface_datasets": 200,
|
| 3 |
+
"huggingface_successful": 12,
|
| 4 |
+
"parallel_datasets": 84,
|
| 5 |
+
"parallel_successful": 0,
|
| 6 |
+
"parallel_total_rows": 0,
|
| 7 |
+
"cc100_oscar_datasets": 2,
|
| 8 |
+
"cc100_oscar_successful": 0,
|
| 9 |
+
"masakhane_datasets": 27,
|
| 10 |
+
"masakhane_successful": 0,
|
| 11 |
+
"wikipedia_articles": 0,
|
| 12 |
+
"igihe_articles": 0,
|
| 13 |
+
"kigalitoday_articles": 0,
|
| 14 |
+
"newtimes_rss_articles": 0,
|
| 15 |
+
"wikisource_texts": 0,
|
| 16 |
+
"wiktionary_entries": 58,
|
| 17 |
+
"rbc_articles": 0,
|
| 18 |
+
"processing": {
|
| 19 |
+
"total_loaded": 1239240,
|
| 20 |
+
"total_after_lang_filter": 755703,
|
| 21 |
+
"total_after_dedup": 661014,
|
| 22 |
+
"total_tokens": 24704887,
|
| 23 |
+
"total_training_samples": 661008,
|
| 24 |
+
"total_instruction": 95437,
|
| 25 |
+
"total_chat": 0,
|
| 26 |
+
"per_file_stats": [
|
| 27 |
+
{
|
| 28 |
+
"file": "huggingface_ChrisToukmaji_kinyarwanda_instruction_tuning.jsonl",
|
| 29 |
+
"loaded": 4250,
|
| 30 |
+
"lang_filtered": 2,
|
| 31 |
+
"clean_filtered": 0,
|
| 32 |
+
"kept": 4248,
|
| 33 |
+
"tokens": 327390,
|
| 34 |
+
"instruction_samples": 4248,
|
| 35 |
+
"chat_samples": 0
|
| 36 |
+
},
|
| 37 |
+
{
|
| 38 |
+
"file": "huggingface_CircuitNotion_kinyarwanda_corpus.jsonl",
|
| 39 |
+
"loaded": 0,
|
| 40 |
+
"lang_filtered": 0,
|
| 41 |
+
"clean_filtered": 0,
|
| 42 |
+
"kept": 0,
|
| 43 |
+
"tokens": 0,
|
| 44 |
+
"instruction_samples": 0,
|
| 45 |
+
"chat_samples": 0
|
| 46 |
+
},
|
| 47 |
+
{
|
| 48 |
+
"file": "huggingface_DigitalUmuganda_kinyarwanda-english-machine-translation-dataset.jsonl",
|
| 49 |
+
"loaded": 0,
|
| 50 |
+
"lang_filtered": 0,
|
| 51 |
+
"clean_filtered": 0,
|
| 52 |
+
"kept": 0,
|
| 53 |
+
"tokens": 0,
|
| 54 |
+
"instruction_samples": 0,
|
| 55 |
+
"chat_samples": 0
|
| 56 |
+
},
|
| 57 |
+
{
|
| 58 |
+
"file": "huggingface_DigitalUmuganda_NMT_Health_parallel_data_en_kin.jsonl",
|
| 59 |
+
"loaded": 30000,
|
| 60 |
+
"lang_filtered": 1,
|
| 61 |
+
"clean_filtered": 4274,
|
| 62 |
+
"kept": 25725,
|
| 63 |
+
"tokens": 524882,
|
| 64 |
+
"instruction_samples": 0,
|
| 65 |
+
"chat_samples": 0
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"file": "huggingface_DigitalUmuganda_NMT_Rwandan-Gazette_parallel_data_en_kin.jsonl",
|
| 69 |
+
"loaded": 80000,
|
| 70 |
+
"lang_filtered": 0,
|
| 71 |
+
"clean_filtered": 9768,
|
| 72 |
+
"kept": 70232,
|
| 73 |
+
"tokens": 1727805,
|
| 74 |
+
"instruction_samples": 0,
|
| 75 |
+
"chat_samples": 0
|
| 76 |
+
},
|
| 77 |
+
{
|
| 78 |
+
"file": "huggingface_jojo-ai-mst_Roleplay-Kinyarwanda.jsonl",
|
| 79 |
+
"loaded": 1923,
|
| 80 |
+
"lang_filtered": 0,
|
| 81 |
+
"clean_filtered": 0,
|
| 82 |
+
"kept": 1923,
|
| 83 |
+
"tokens": 50283,
|
| 84 |
+
"instruction_samples": 1923,
|
| 85 |
+
"chat_samples": 0
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"file": "huggingface_mbazaNLP_Kinyarwanda_English_parallel_dataset.jsonl",
|
| 89 |
+
"loaded": 55667,
|
| 90 |
+
"lang_filtered": 0,
|
| 91 |
+
"clean_filtered": 7384,
|
| 92 |
+
"kept": 48283,
|
| 93 |
+
"tokens": 779663,
|
| 94 |
+
"instruction_samples": 0,
|
| 95 |
+
"chat_samples": 0
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"file": "huggingface_michsethowusu_akan-kinyarwanda_sentence-pairs.jsonl",
|
| 99 |
+
"loaded": 67145,
|
| 100 |
+
"lang_filtered": 7800,
|
| 101 |
+
"clean_filtered": 39704,
|
| 102 |
+
"kept": 19641,
|
| 103 |
+
"tokens": 294505,
|
| 104 |
+
"instruction_samples": 0,
|
| 105 |
+
"chat_samples": 0
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"file": "huggingface_michsethowusu_bambara-kinyarwanda_sentence-pairs.jsonl",
|
| 109 |
+
"loaded": 56301,
|
| 110 |
+
"lang_filtered": 3366,
|
| 111 |
+
"clean_filtered": 25937,
|
| 112 |
+
"kept": 26998,
|
| 113 |
+
"tokens": 522282,
|
| 114 |
+
"instruction_samples": 0,
|
| 115 |
+
"chat_samples": 0
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"file": "huggingface_michsethowusu_bemba-kinyarwanda_sentence-pairs.jsonl",
|
| 119 |
+
"loaded": 197886,
|
| 120 |
+
"lang_filtered": 2,
|
| 121 |
+
"clean_filtered": 101878,
|
| 122 |
+
"kept": 96006,
|
| 123 |
+
"tokens": 1379707,
|
| 124 |
+
"instruction_samples": 0,
|
| 125 |
+
"chat_samples": 0
|
| 126 |
+
},
|
| 127 |
+
{
|
| 128 |
+
"file": "huggingface_michsethowusu_dinka-kinyarwanda_sentence-pairs.jsonl",
|
| 129 |
+
"loaded": 39196,
|
| 130 |
+
"lang_filtered": 6663,
|
| 131 |
+
"clean_filtered": 18172,
|
| 132 |
+
"kept": 14361,
|
| 133 |
+
"tokens": 268754,
|
| 134 |
+
"instruction_samples": 0,
|
| 135 |
+
"chat_samples": 0
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"file": "huggingface_michsethowusu_dyula-kinyarwanda_sentence-pairs.jsonl",
|
| 139 |
+
"loaded": 102925,
|
| 140 |
+
"lang_filtered": 2446,
|
| 141 |
+
"clean_filtered": 57811,
|
| 142 |
+
"kept": 42668,
|
| 143 |
+
"tokens": 857116,
|
| 144 |
+
"instruction_samples": 0,
|
| 145 |
+
"chat_samples": 0
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"file": "huggingface_michsethowusu_kinyarwanda-rundi_sentence-pairs.jsonl",
|
| 149 |
+
"loaded": 410000,
|
| 150 |
+
"lang_filtered": 11,
|
| 151 |
+
"clean_filtered": 163087,
|
| 152 |
+
"kept": 246902,
|
| 153 |
+
"tokens": 3516163,
|
| 154 |
+
"instruction_samples": 0,
|
| 155 |
+
"chat_samples": 0
|
| 156 |
+
},
|
| 157 |
+
{
|
| 158 |
+
"file": "huggingface_Mikecyane_Kinyarwanda_chat.jsonl",
|
| 159 |
+
"loaded": 0,
|
| 160 |
+
"lang_filtered": 0,
|
| 161 |
+
"clean_filtered": 0,
|
| 162 |
+
"kept": 0,
|
| 163 |
+
"tokens": 0,
|
| 164 |
+
"instruction_samples": 0,
|
| 165 |
+
"chat_samples": 0
|
| 166 |
+
},
|
| 167 |
+
{
|
| 168 |
+
"file": "huggingface_RogerB_kin_en_DigitalUmuganda.jsonl",
|
| 169 |
+
"loaded": 47824,
|
| 170 |
+
"lang_filtered": 0,
|
| 171 |
+
"clean_filtered": 30920,
|
| 172 |
+
"kept": 16904,
|
| 173 |
+
"tokens": 159881,
|
| 174 |
+
"instruction_samples": 0,
|
| 175 |
+
"chat_samples": 0
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"file": "huggingface_saillab_alpaca-kinyarwanda-cleaned.jsonl",
|
| 179 |
+
"loaded": 41601,
|
| 180 |
+
"lang_filtered": 2,
|
| 181 |
+
"clean_filtered": 1900,
|
| 182 |
+
"kept": 39699,
|
| 183 |
+
"tokens": 1440835,
|
| 184 |
+
"instruction_samples": 41599,
|
| 185 |
+
"chat_samples": 0
|
| 186 |
+
},
|
| 187 |
+
{
|
| 188 |
+
"file": "huggingface_saillab_alpaca_kinyarwanda_taco.jsonl",
|
| 189 |
+
"loaded": 49601,
|
| 190 |
+
"lang_filtered": 11,
|
| 191 |
+
"clean_filtered": 0,
|
| 192 |
+
"kept": 49590,
|
| 193 |
+
"tokens": 12116090,
|
| 194 |
+
"instruction_samples": 49590,
|
| 195 |
+
"chat_samples": 0
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"file": "huggingface_souvorinkg_english_kinyarwanda.jsonl",
|
| 199 |
+
"loaded": 54589,
|
| 200 |
+
"lang_filtered": 0,
|
| 201 |
+
"clean_filtered": 2387,
|
| 202 |
+
"kept": 52202,
|
| 203 |
+
"tokens": 662665,
|
| 204 |
+
"instruction_samples": 0,
|
| 205 |
+
"chat_samples": 0
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"file": "igihe_igihe_articles.jsonl",
|
| 209 |
+
"loaded": 0,
|
| 210 |
+
"lang_filtered": 0,
|
| 211 |
+
"clean_filtered": 0,
|
| 212 |
+
"kept": 0,
|
| 213 |
+
"tokens": 0,
|
| 214 |
+
"instruction_samples": 0,
|
| 215 |
+
"chat_samples": 0
|
| 216 |
+
},
|
| 217 |
+
{
|
| 218 |
+
"file": "kigalitoday_kigalitoday_articles.jsonl",
|
| 219 |
+
"loaded": 0,
|
| 220 |
+
"lang_filtered": 0,
|
| 221 |
+
"clean_filtered": 0,
|
| 222 |
+
"kept": 0,
|
| 223 |
+
"tokens": 0,
|
| 224 |
+
"instruction_samples": 0,
|
| 225 |
+
"chat_samples": 0
|
| 226 |
+
},
|
| 227 |
+
{
|
| 228 |
+
"file": "newtimes_rss_newtimes_rss_articles.jsonl",
|
| 229 |
+
"loaded": 0,
|
| 230 |
+
"lang_filtered": 0,
|
| 231 |
+
"clean_filtered": 0,
|
| 232 |
+
"kept": 0,
|
| 233 |
+
"tokens": 0,
|
| 234 |
+
"instruction_samples": 0,
|
| 235 |
+
"chat_samples": 0
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"file": "rbc_rbc_articles.jsonl",
|
| 239 |
+
"loaded": 0,
|
| 240 |
+
"lang_filtered": 0,
|
| 241 |
+
"clean_filtered": 0,
|
| 242 |
+
"kept": 0,
|
| 243 |
+
"tokens": 0,
|
| 244 |
+
"instruction_samples": 0,
|
| 245 |
+
"chat_samples": 0
|
| 246 |
+
},
|
| 247 |
+
{
|
| 248 |
+
"file": "wikipedia_wikipedia_articles.jsonl",
|
| 249 |
+
"loaded": 274,
|
| 250 |
+
"lang_filtered": 0,
|
| 251 |
+
"clean_filtered": 0,
|
| 252 |
+
"kept": 274,
|
| 253 |
+
"tokens": 75991,
|
| 254 |
+
"instruction_samples": 0,
|
| 255 |
+
"chat_samples": 0
|
| 256 |
+
},
|
| 257 |
+
{
|
| 258 |
+
"file": "wikisource_wikisource_articles.jsonl",
|
| 259 |
+
"loaded": 0,
|
| 260 |
+
"lang_filtered": 0,
|
| 261 |
+
"clean_filtered": 0,
|
| 262 |
+
"kept": 0,
|
| 263 |
+
"tokens": 0,
|
| 264 |
+
"instruction_samples": 0,
|
| 265 |
+
"chat_samples": 0
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"file": "wiktionary_wiktionary_entries.jsonl",
|
| 269 |
+
"loaded": 58,
|
| 270 |
+
"lang_filtered": 11,
|
| 271 |
+
"clean_filtered": 0,
|
| 272 |
+
"kept": 47,
|
| 273 |
+
"tokens": 875,
|
| 274 |
+
"instruction_samples": 0,
|
| 275 |
+
"chat_samples": 0
|
| 276 |
+
}
|
| 277 |
+
],
|
| 278 |
+
"output_files": {
|
| 279 |
+
"training": "data\\processed\\training_data.jsonl",
|
| 280 |
+
"instruction": "data\\processed\\instruction_data.jsonl",
|
| 281 |
+
"chat": "data\\processed\\chat_data.jsonl"
|
| 282 |
+
}
|
| 283 |
+
}
|
| 284 |
+
}
|