Dataset Viewer
Auto-converted to Parquet Duplicate
source
stringclasses
6 values
weight
float64
0.05
0.45
sequences
int64
9.77M
87.9M
training_tokens
int64
10B
90B
source_tokens_consumed
int64
10B
90B
sequence_length
int64
1.02k
1.02k
token_dtype
stringclasses
1 value
shards
int64
11
91
cosmopedia_v2
0.05
9,765,632
10,000,007,168
10,000,007,169
1,024
uint16
11
dclm
0.45
87,890,688
90,000,064,512
90,000,064,513
1,024
uint16
91
finemath
0.05
9,765,632
10,000,007,168
10,000,007,169
1,024
uint16
11
fineweb_edu_dedup
0.3
58,593,792
60,000,043,008
60,000,043,009
1,024
uint16
61
infiwebmath
0.05
9,765,632
10,000,007,168
10,000,007,169
1,024
uint16
11
stack_edu
0.1
19,531,264
20,000,014,336
20,000,014,337
1,024
uint16
21

TR-HASH 32K · 200B Token Mixture

Pretokenized training mixture for compact TR-HASH language-model research.

The Dataset Viewer displays one summary row per source. The actual training data is stored as packed uint16 token shards under corpora/<source>/tokens-*.bin. Each sequence contains 1,024 token IDs produced by the project 32K tokenizer.

Mixture

Source Weight Training tokens
DCLM 45% 90B
FineWeb-Edu deduplicated 30% 60B
Stack-Edu 10% 20B
FineMath 5% 10B
InfiWebMath 5% 10B
Cosmopedia v2 5% 10B

Technical manifests are stored beside each corpus. mixture_manifest.json describes the complete mixture and is not part of the Viewer split.

Downloads last month
256