Foundry-ML simplifies access to machine learning-ready datasets in materials science and chemistry.
- Search & Load - Find and use curated datasets with a few lines of code
- Understand - Rich schemas describe what each field means
- Cite - Automatic citation generation for publications
- Publish - Share your datasets with the community
- AI-Ready - MCP server for Claude and other AI assistants
Quick Start
pip install foundry-ml
Need optional integrations? Install extras only when you need them:
pip install "foundry-ml[torch]" # Enable dataset.get_as_torch()
pip install "foundry-ml[tensorflow]" # Enable dataset.get_as_tensorflow()
pip install "foundry-ml[huggingface]" # Enable push-to-hub
pip install "foundry-ml[excel]" # Excel import support via openpyxl
PyTorch/TensorFlow extras expect wheels compiled against NumPy 2.0. Install PyTorch 2.3+ and TensorFlow 2.18+ (or newer builds with NumPy 2 support) to avoid ABI errors.
from foundry import Foundry
# Connect
f = Foundry()
# Search
results = f.search("band gap", limit=5)
# Load
dataset = results.iloc[0].FoundryDataset
X, y = dataset.get_as_dict()['train']
# Understand
schema = dataset.get_schema()
print(schema['fields'])
# Cite
print(dataset.get_citation())
Cloud Environments
For Google Colab or remote Jupyter:
f = Foundry(no_browser=True, no_local_server=True)
CLI
foundry search "band gap"
foundry schema 10.18126/abc123
foundry --help
AI Agent Integration
foundry mcp install # Add to Claude Code
Documentation
Features
| Feature | Description |
|---|---|
| Search | Find datasets by keyword, DOI, or browse catalog |
| Load | Automatic download, caching, and format conversion |
| PyTorch/TensorFlow (extras) | dataset.get_as_torch(), dataset.get_as_tensorflow() |
| CLI | Terminal-based workflows |
| MCP Server | AI assistant integration |
| HuggingFace Export (extra) | Publish to HuggingFace Hub |
Available Datasets
Browse datasets at Foundry-ML.org or:
f = Foundry()
f.list(limit=20) # See available datasets
How to Cite
If you use Foundry-ML, please cite:
@article{Schmidt2024,
doi = {10.21105/joss.05467},
year = {2024},
publisher = {The Open Journal},
volume = {9},
number = {93},
pages = {5467},
author = {Kj Schmidt and Aristana Scourtas and Logan Ward and others},
title = {Foundry-ML - Software and Services to Simplify Access to Machine Learning Datasets in Materials Science},
journal = {Journal of Open Source Software}
}
Contributing
Foundry is open source. To contribute:
- Fork from
main - Make your changes
- Open a Pull Request
See CONTRIBUTING.md for details.
Support
This work was supported by the National Science Foundation under NSF Award Number: 1931306 "Collaborative Research: Framework: Machine Learning Materials Innovation Infrastructure".
Foundry integrates with Materials Data Facility, FuncX, and MAST-ML.
Release files for foundry-ml 1.2.2
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| foundry_ml-1.2.2.tar.gz | 55.5 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| foundry_ml-1.2.2-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 119.3 kB
Release files / foundry_ml-1.2.2.tar.gz
| Download URL | foundry_ml-1.2.2.tar.gz |
|---|---|
| Size | 55.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
cbb8f26c70cd17b17fbb78f4eecf55ee3ea543ab658934b6f120b57ca0ef4c36
|
|
BLAKE2b-256 checksum How to use checksums |
3b99b5eb09caf8fad27b2a7821fc42d149dbb9b737730bb5fb56f32fa7be6169
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.14.2
|
Release files / foundry_ml-1.2.2-py3-none-any.whl
| Download URL | foundry_ml-1.2.2-py3-none-any.whl |
|---|---|
| Size | 63.8 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
22fcd641e11f97e27c483eeb6f8489adc0675c0d487e419495baadfe6c228a7c
|
|
BLAKE2b-256 checksum How to use checksums |
d353bf02798c8c675d08386a77b41046ec957419ebe46f432b0547a6004f4c39
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.14.2
|
