ai2-adapt-dev/flan_v2_converted
This is a converted version of the Flan dataset into Tulu SFT training format. The conversion script can be found in our open-instruct repo. The conversion took the following parameters: apply_keyword_filters: True apply_empty_message_filters: True push_to_hub: True hf_entity: ai2-adapt-dev converted_dataset_name: flan_v2_converted local_save_dir: ./data/sft/flan The original FLAN dataset needs extensive efforts to be regenerated, so we are using a reproduced version by the OpenOrca… See the full description on the dataset page: https://huggingface.co/datasets/ai2-adapt-dev/flan_v2_converted.
| Author / organization | ai2-adapt-dev |
|---|---|
| Downloads | 178 (as reported by Hugging Face) |
| Likes | 3 |
| Created | 14 October 2024 |
| Last updated at source | 14 October 2024 |
Popularity, last 2 days
178 now (-12 since 2026-10-16). Counts as reported by the source, snapshotted daily by Silytics.
Connected in the knowledge graph
- mradermacher/Q3-30B-A3B-Designant-i1-GGUFModel · used to train
- mradermacher/Q3-8B-Kintsugi-GGUFModel · used to train
- OpenLLM-France/Lucie-7B-Instruct-v1.1Model · used to train
- bartowski/allura-org_Q3-30B-A3B-Designant-GGUFModel · used to train
- mradermacher/Lucie-7B-Instruct-v1.1-GGUFModel · used to train
- tensorblock/Lucie-7B-Instruct-v1.1-GGUFModel · used to train
- mradermacher/Q3-30B-A3B-Designant-GGUFModel · used to train
- allura-quants/allura-org_MS3.2-24b-Angel-GGUFModel · used to train
- OpenLLM-France/Lucie-7B-Instruct-human-dataModel · used to train
- mradermacher/MS3.2-24b-Angel-GGUFModel · used to train
- mradermacher/MS3.2-24b-Angel-i1-GGUFModel · used to train
- mradermacher/Lucie-7B-Instruct-human-data-GGUFModel · used to train
- allura-org/Q3-30B-A3B-DesignantModel · used to train
- mradermacher/Q3-8B-Kintsugi-i1-GGUFModel · used to train
More from ai2-adapt-dev
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.