Hugging Face Hub для научного ML

★ 7.0 · learning

hugging-science is a Claude Code skill that helps discover and use scientific datasets, models, blog posts, and interactive demos from Hugging Face Hub for ML work across 17 domains — including genomics, biology, chemistry, climate, medicine, protein design, and drug discovery. The catalog at huggingscience.co provides a structured, LLM-readable index via `llms.txt`, `llms-full.txt`, and per-domain markdown files, while the `hugging-science` Hugging Face organization hosts around 27 interactive Spaces such as BoltzGen for protein and binder design. Resources are accessed through standard Hugging Face APIs: the `datasets` and `transformers` libraries, the Inference API, and `gradio_client` for Spaces. Researchers looking to fine-tune on scientific data, reproduce a scientific ML paper, or find the right model — such as ESM2 for protein sequences or Nucleotide Transformer for genomics — will find the pre-filtered, quality-curated catalog far more efficient than a generic Hub search.