Long-context (16K packing) pretraining releases, same two-parquet contract as the regular mixes. Token counts on each card.
-
placeholderlabs/pretrain-repository-v2-mix-long-context
Viewer • Updated • 5.16M • 89 -
placeholderlabs/pretrain-academic-mix-long-context
Viewer • Updated • 2M • 160 -
placeholderlabs/pretrain-commits-v2-mix-long-context
Viewer • Updated • 1.98M • 348 -
placeholderlabs/pretrain-web-mix-long-context
Viewer • Updated • 564k • 428