Dataset Viewer
Auto-converted to Parquet Duplicate
Search is not available for this dataset
input_ids
listlengths
4.1k
4.1k
[ 304, 16327, 753, 386, 4428, 480, 33312, 1614, 11, 374, 2162, 311, 264, 44797, 32726, 25, 20543, 27840, 67487, 320, 36793, 700, 64, 389, 936, 570, 763, 264, 13945, 4756, 4007, 11, 264, 1912, 315, 6625, 12074, 9477, 430, 279, 6913, 1017...
[ 323, 264, 30104, 33315, 520, 279, 3907, 315, 19174, 11, 323, 3010, 520, 19326, 3907, 13, 1283, 574, 264, 4562, 315, 279, 5250, 370, 1094, 23488, 8925, 315, 39898, 1105, 505, 220, 2550, 20, 311, 220, 1049, 17, 13, 1283, 8636, 304, 22...
[ 9349, 13, 2030, 420, 27373, 13340, 6724, 27227, 279, 5150, 28970, 315, 279, 3278, 311, 89959, 872, 1866, 3917, 1968, 304, 279, 3663, 315, 8987, 21448, 627, 10714, 279, 7030, 198, 30700, 6729, 31930, 16732, 73268, 574, 264, 17054, 315, 6...
[ 364, 2170, 45752, 6, 323, 617, 6575, 2653, 311, 1893, 369, 5694, 264, 21955, 4029, 2930, 323, 304, 279, 1972, 1917, 2001, 323, 279, 80267, 990, 1912, 15849, 430, 7000, 315, 420, 374, 3062, 3403, 311, 2567, 872, 836, 13989, 13, 2435, ...
[ 1093, 279, 650, 1509, 323, 279, 452, 2836, 26, 323, 433, 374, 10171, 430, 96812, 4311, 9076, 439, 28235, 304, 279, 16566, 70684, 460, 22260, 13, 578, 2057, 263, 2979, 574, 2500, 96812, 9932, 304, 1521, 87733, 26, 279, 36769, 5808, 574...
[ 574, 832, 31394, 990, 315, 10748, 6761, 627, 3112, 568, 1903, 813, 8254, 50552, 11, 323, 813, 4224, 20964, 11, 323, 813, 4224, 1386, 67, 21168, 11, 315, 10748, 6761, 627, 2173, 264, 11005, 315, 10748, 6761, 1903, 568, 433, 11, 323, ...
[ 10390, 5814, 320, 14938, 38, 9483, 1372, 8, 36201, 649, 387, 1511, 311, 7417, 279, 74633, 3210, 323, 76041, 315, 54626, 1521, 63324, 706, 1027, 279, 3917, 315, 3892, 26969, 627, 791, 3560, 315, 54736, 437, 8728, 304, 279, 6156, 16865, ...
[ 82, 31873, 2476, 279, 2324, 315, 22706, 389, 279, 11841, 490, 12055, 682, 1023, 41329, 10742, 13, 3277, 1701, 279, 4751, 11, 41329, 28424, 1288, 5357, 389, 9045, 311, 14201, 279, 11841, 1555, 57340, 11919, 12659, 430, 649, 8854, 22706, ...
[ 5675, 77324, 3626, 627, 8468, 220, 18, 13, 32341, 323, 11993, 5675, 77324, 3626, 627, 6153, 36201, 11, 1120, 17562, 682, 11993, 481, 3626, 13, 5112, 3373, 279, 77324, 3626, 323, 3665, 1124, 389, 701, 2653, 6678, 627, 5519, 4718, 2956, ...
[ 296, 3935, 73, 19807, 320, 44, 3935, 18029, 355, 296, 3935, 63502, 705, 8545, 712, 1590, 320, 78524, 1156, 299, 3716, 705, 323, 1524, 436, 17082, 355, 9155, 20234, 320, 20122, 17544, 15975, 33019, 570, 1102, 374, 13240, 430, 1070, 527, ...
End of preview. Expand in Data Studio

MBT Validation

The fixed validation sequences used for Multi-Branch Transformer. This release preserves the original token IDs and sample order, allowing evaluation without rebuilding dataset indexes or downloading the full corpus.

Split Rows Column Tokens per row Evaluated targets
validation 12,160 input_ids: list of int32 4,097 49,807,360

Each row contains 4,096 input tokens plus one additional next-token target. All targets contribute to the loss, so no loss-mask column is needed. Preserve row order and token IDs; do not retokenize or add special tokens.

from datasets import load_dataset

data = load_dataset("gaotang/MBT-Validation", split="validation")
row = data[0]["input_ids"]
inputs = row[:-1]
targets = row[1:]

Use causal attention and positions 0 through 4095 within each sequence. Neither positions nor attention reset at document boundaries. Compute validation loss as the sum of next-token cross-entropies divided by the number of evaluated targets; perplexity is its exponential.

Source and preparation

The source is FineWeb-Edu, configuration sample-100BT, at revision 87f09149ef4734204d70ed1d046ddc9ca3f2b8f9. We held out sample/100BT/000_00000.parquet from MBT pretraining. This is our held-out subset, not an official upstream validation split.

Documents were tokenized with the Llama 3 tokenizer, ended with EOD token 128001, and packed into sequences. The release contains the first 12,160 samples in the original validation order generated with seed 1234. The vocabulary contains 128,256 token IDs; tokenizer files are not included.

FineWeb-Edu is provided by Hugging Face under ODC-By. This derived dataset retains that attribution and license.

Downloads last month
14

Paper for gaotang/MBT-Validation