| # Destructive Sampling Report |
|
|
| - tokenized_path: `/nemo-workspace/inf-evolve/home/project/HRM-Text/data_io/data_tokenized_bpe_65k` |
| - output_path: `/nemo-workspace/inf-evolve/home/project/HRM-Text/data/sampled` |
| - epochs: `4` |
| - context_size: `4097` |
| - total_length: `14980217200` |
| - source token files deleted: `True` |
|
|
| | Group | Tasks | Rows | Sampled rows/epoch | Tokens after filter | |
| |---|---:|---:|---:|---:| |
| | Platypus | 10 | 33,752 | 337,520 | 8,303,655 | |
| | SYNTH | 500 | 60,789,130 | 10,000,000 | 21,692,446,150 | |
| | acereason | 1 | 2,668,708 | 2,000,000 | 1,869,256,755 | |
| | amps_khan.jsonl | 1 | 103,059 | 103,059 | 30,333,670 | |
| | ampsmathematica | 137 | 4,830,035 | 1,210,500 | 681,051,184 | |
| | dmmath | 168 | 111,999,888 | 16,800,000 | 4,899,383,195 | |
| | flan | 4,208 | 377,563,081 | 14,481,869 | 138,385,858,380 | |
| | gsm8k_train.jsonl | 1 | 7,473 | 74,730 | 539,265 | |
| | math_train.jsonl | 1 | 14,996 | 149,960 | 2,956,066 | |
| | natural_reasoning.jsonl | 1 | 770,141 | 770,141 | 101,371,244 | |
| | no_robots.jsonl | 1 | 10,000 | 100,000 | 3,085,450 | |
| | numinamath.jsonl | 1 | 811,919 | 811,919 | 249,871,208 | |
| | omnimath.jsonl | 1 | 8,847 | 88,470 | 2,717,694 | |
| | openmathinstruct2 | 2 | 25,020,108 | 4,000,000 | 6,550,216,284 | |
| | openthoughts2 | 1 | 892,157 | 500,000 | 519,094,835 | |
| | principia_collection.jsonl | 1 | 554,399 | 554,399 | 233,161,445 | |
| | sudoku_extreme | 1 | 3,831,994 | 1,000,000 | 662,934,962 | |
| | tasksource | 182 | 2,363,540 | 1,251,675 | 306,044,246 | |
| | textbookreasoning | 2 | 1,178,012 | 1,178,012 | 358,146,301 | |
| | webinstruct_verified.jsonl | 1 | 228,726 | 2,287,260 | 23,840,853 | |
|
|
