metadata
license: apache-2.0
task_categories:
- text-generation
language:
- en
configs:
- config_name: default
data_files:
- split: train
path:
- shard_00000.parquet
- shard_00001.parquet
- shard_00002.parquet
- shard_00003.parquet
- shard_00004.parquet
- shard_00005.parquet
- shard_00006.parquet
- shard_00007.parquet
- shard_00008.parquet
- shard_00009.parquet
- shard_00010.parquet
- shard_00011.parquet
- shard_00012.parquet
- shard_00013.parquet
- shard_00014.parquet
- shard_00015.parquet
- shard_00016.parquet
- shard_00017.parquet
- shard_00018.parquet
- shard_00019.parquet
- shard_00020.parquet
- shard_00021.parquet
- shard_00022.parquet
- shard_00023.parquet
- shard_00024.parquet
- shard_00025.parquet
- shard_00026.parquet
- shard_00027.parquet
- shard_00028.parquet
- shard_00029.parquet
- shard_00030.parquet
- shard_00031.parquet
- shard_00032.parquet
- shard_00033.parquet
- shard_00034.parquet
- shard_00035.parquet
- shard_00036.parquet
- shard_00037.parquet
- shard_00038.parquet
- shard_00039.parquet
- shard_00040.parquet
- shard_00041.parquet
- shard_00042.parquet
- shard_00043.parquet
- shard_00044.parquet
- shard_00045.parquet
- shard_00046.parquet
- shard_00047.parquet
- shard_00048.parquet
- shard_00049.parquet
- shard_00050.parquet
- shard_00051.parquet
- shard_00052.parquet
- shard_00053.parquet
- shard_00054.parquet
- shard_00055.parquet
- shard_00056.parquet
- shard_00057.parquet
- shard_00058.parquet
- shard_00059.parquet
Dataset
Mixed pretraining dataset built from:
| Source | Weight | Notes |
|---|---|---|
| epfml/FineWeb-HQ | 60% | Quality-filtered web text (score > 0.8) |
| HuggingFaceTB/cosmopedia | 20% | Sequential: stanford → wikihow → web_samples_v2 |
| HuggingFaceTB/finemath (finemath-4plus) | 10% | Mathematical reasoning |
| bigcode/python-stack-v1-functions-filtered | 10% | Python code functions |
Total rows: 6,000,000
Shards: 60 parquet files
Split: all rows are in train