SKU SFT-2106 · Sold by External
FLAN Collection (Flan v2)
Product specifications
| SKU | SFT-2106 |
|---|---|
| Data type | SFT dataset |
| Volume | examples |
| Size on disk | Not published by source |
| Format | Gzip-compressed JSONL (Parquet auto-conversion available) |
| Access model | PUBLIC LICENSE |
| Pricing | Free · open-source license |
| Quality score | — |
| License | Apache-2.0 |
SirNeural/flan_v2 is an easier-to-access re-release of Google Research's Flan 2022 Collection, aggregating five task families (flan, t0, niv2, cot, dialog) each in zero-shot/few-shot × option/no-option variants (20 submixes; 1,836 templated tasks). Rows are simple inputs/targets/task triples stored as gzip-compressed JSONL. The full collection is very large — the HF datasets-server estimates ~336M rows total, though the auto-converted Parquet view exposes ~2.46M; treat the exact headline count as unverified.