MolAS-Active-Learning / docs /release_workflow.md
SeanWan05's picture
Publish MolAS Active Learning AI4DD dataset and research artifacts
84a242a verified
|
Raw History Blame Contribute Delete
853 Bytes

Hugging Face release workflow

The complete local upload payload includes the graph files and is ready for the chosen dataset repository. Follow UPLOAD.md. The GitHub code-release records are archived under provenance/github_snapshot; they describe the earlier code publication, not current HF file exclusions.

Validate with python3 scripts/verify_hf_package.py and python3 scripts/validate_data.py --require-complete. The former checks the complete HF manifest, including graphs. No scientific runs are required for upload.

The source reference is recorded in provenance/hf_source_snapshot.json. Author-owned data use CC BY 4.0, original software uses MIT, and third-party components retain their terms. After uploading, verify actual public access and the dataset revision before adding the URL to the paper and GitHub README.