ChemPile is a multimodal chemistry corpus containing text, molecular captions, images, code, papers, and structured chemical records. This repository provides eight original components, totaling about 193 GB, for chemistry language and multimodal modeling.

Part of the Molecular Datasets collection.

Get started

Original files are available in their native formats.

Where the data comes from

Files and modalities

Modality	Where	Content
source files	files/caption/	12 original files; 3.23 GB
source files	files/code/	91 original files; 15.64 GB
source files	files/education/	15 original files; 0.26 GB
source files	files/instruction/	14 original files; 0.61 GB
source files	files/lift/	503 original files; 18.49 GB
source files	files/mlift/	1,809 original files; 123.40 GB
source files	files/paper/	145 original files; 31.58 GB
source files	files/reasoning/	66 original files; 0.12 GB

Download an individual file:

from huggingface_hub import hf_hub_download
path = hf_hub_download("liuganghuggingface/moldata-chempile", "files/caption/data/test-00000-of-00001.parquet", repo_type="dataset")

Dataset notes

Component licenses are listed individually. The paper component uses a no-derivatives license.
Image/text pairs follow each component’s record relationships; components use separate identifier systems.
The corpus contains heterogeneous chemistry data rather than one universal SMILES/property prediction task.

License

Mixed component licenses: CC BY-NC-SA 4.0; CC BY-NC-ND 4.0; CC BY-SA 4.0; CC BY 4.0; AGPL-3.0 and original per-record code licenses

Component	License	Source
education	cc-by-nc-sa-4.0	Original component
paper	cc-by-nc-nd-4.0	Original component
mlift	cc-by-nc-sa-4.0	Original component
reasoning	cc-by-sa-4.0	Original component
code	agpl-3.0	Original component
caption	cc-by-nc-sa-4.0	Original component
lift	cc-by-nc-sa-4.0	Original component
instruction	cc-by-4.0	Original component

Citation

Cite the original study and data source:

Downloads last month: 28

Total file size:

193 GB

Collection including liuganghuggingface/moldata-chempile

Molecular Datasets

Collection

Small-molecule and polymer datasets. Start from molecular-datasets-index. https://liugangcode.github.io/molecular-dataset-hub/ • 100 items • Updated 1 day ago