SmolTalk
← All datasets
Author: Distillio · License: mixed (apache-2.0 + source licenses) · Format: parquet
0 downloads · 0 stars · 0 mirrors · size unknown · 1 file
Curated Magpie-Ultra-based instruction mix used to train the SmolLM2 family.
Tags: sft · instruction-tuning · magpie
Description
Curated instruction-tuning mix built around Magpie-Ultra, used to train the SmolLM2 family. A mix: newly-generated subsets are Apache-2.0, but included public subsets keep their original licenses.
Download
Direct download: https://huggingface.co/datasets/HuggingFaceTB/smoltalk/resolve/main/data/smol-magpie-ultra/train-00001-of-00006.parquet
Or via the API: curl -L -o file "https://huggingface.co/datasets/HuggingFaceTB/smoltalk/resolve/main/data/smol-magpie-ultra/train-00001-of-00006.parquet"
Files in the Hugging Face repository
- .gitattributes (2.4 KB)
- data/all/.gitattributes (2.4 KB)
- data/all/test-00000-of-00001.parquet (101 MB)
- data/all/train-00000-of-00009.parquet (213 MB)
- data/all/train-00001-of-00009.parquet (213 MB)
- data/all/train-00002-of-00009.parquet (213 MB)
- data/all/train-00003-of-00009.parquet (213 MB)
- data/all/train-00004-of-00009.parquet (213 MB)
- data/all/train-00005-of-00009.parquet (212 MB)
- data/all/train-00006-of-00009.parquet (212 MB)
- data/all/train-00007-of-00009.parquet (214 MB)
- data/all/train-00008-of-00009.parquet (214 MB)
- data/apigen-80k/.gitattributes (2.4 KB)
- data/apigen-80k/test-00000-of-00001.parquet (2.5 MB)
- data/apigen-80k/train-00000-of-00001.parquet (47.5 MB)
- data/everyday-conversations/.gitattributes (2.4 KB)
- data/everyday-conversations/test-00000-of-00001.parquet (51.4 KB)
- data/everyday-conversations/train-00000-of-00001.parquet (924 KB)
- data/explore-instruct-rewriting/.gitattributes (2.4 KB)
- data/explore-instruct-rewriting/test-00000-of-00001.parquet (279 KB)
- data/explore-instruct-rewriting/train-00000-of-00001.parquet (5.1 MB)
- data/longalign/.gitattributes (2.4 KB)
- data/longalign/test-00000-of-00001.parquet (3.5 MB)
- data/longalign/train-00000-of-00001.parquet (63.4 MB)
- data/metamathqa-50k/.gitattributes (2.4 KB)
- data/metamathqa-50k/test-00000-of-00001.parquet (790 KB)
- data/metamathqa-50k/train-00000-of-00001.parquet (14.8 MB)
- data/numina-cot-100k/.gitattributes (2.4 KB)
- data/numina-cot-100k/test-00000-of-00001.parquet (3.8 MB)
- data/numina-cot-100k/train-00000-of-00001.parquet (70.6 MB)
- data/openhermes-100k/.gitattributes (2.4 KB)
- data/openhermes-100k/test-00000-of-00001.parquet (3.9 MB)
- data/openhermes-100k/train-00000-of-00001.parquet (75.6 MB)
- data/self-oss-instruct/.gitattributes (2.4 KB)
- data/self-oss-instruct/test-00000-of-00001.parquet (1.3 MB)
- data/self-oss-instruct/train-00000-of-00001.parquet (25.0 MB)
- data/smol-constraints/.gitattributes (2.4 KB)
- data/smol-constraints/test-00000-of-00001.parquet (866 KB)
- data/smol-constraints/train-00000-of-00001.parquet (16.1 MB)
- data/smol-magpie-ultra/.gitattributes (2.4 KB)
- data/smol-magpie-ultra/test-00000-of-00001.parquet (69.8 MB)
- data/smol-magpie-ultra/train-00000-of-00006.parquet (222 MB)
- data/smol-magpie-ultra/train-00001-of-00006.parquet (222 MB)
- data/smol-magpie-ultra/train-00002-of-00006.parquet (222 MB)
- data/smol-magpie-ultra/train-00003-of-00006.parquet (222 MB)
- data/smol-magpie-ultra/train-00004-of-00006.parquet (222 MB)
- data/smol-magpie-ultra/train-00005-of-00006.parquet (222 MB)
- data/smol-rewrite/.gitattributes (2.4 KB)
- data/smol-rewrite/test-00000-of-00001.parquet (2.0 MB)
- data/smol-rewrite/train-00000-of-00001.parquet (37.1 MB)
- data/smol-summarize/.gitattributes (2.4 KB)
- data/smol-summarize/test-00000-of-00001.parquet (5.9 MB)
- data/smol-summarize/train-00000-of-00001.parquet (114 MB)
- data/systemchats-30k/.gitattributes (2.4 KB)
- data/systemchats-30k/test-00000-of-00001.parquet (2.4 MB)
- data/systemchats-30k/train-00000-of-00001.parquet (45.7 MB)
- README.md (9.5 KB)
Browse all files in the HuggingFaceTB/smoltalk repository on Hugging Face
Added 2026-08-11. Platform: Distillio — the open source training data collective.