Croissant

Croissant 🥐 is a high-level format building on schema.org for machine learning datasets that combines metadata, resource file descriptions, data structure, and default ML semantics into a single file. It works with existing datasets to make them easier to find, use, and support with tools.

Here, we demonstrate how LaminDB can be used to validate Croissant files:

# pip install lamindb
!lamin init --storage ./test-lamin-croissant

Show code cell output Hide code cell output

→ initialized lamindb: anonymous/test-lamin-croissant

import lamindb as ln
import json

ln.track()

Show code cell output Hide code cell output

→ connected lamindb: anonymous/test-lamin-croissant

→ created Transform('W4Im6iUQV8bK0000', key='croissant.ipynb'), started new Run('RXmr7QcGEwN73f8V') at 2025-11-05 21:34:23 UTC

→ notebook imports: lamindb==1.15.0

• recommendation: to identify the notebook across renames, pass the uid: ln.track("W4Im6iUQV8bK")

croissant_path, dataset1_path = ln.examples.croissant.mini_immuno()
croissant_path

Show code cell output Hide code cell output

PosixPath('mini_immuno.anndata.zarr_metadata.json')

with open(croissant_path) as f:
    dictionary = json.load(f)

print(json.dumps(dictionary, indent=2))

Show code cell output Hide code cell output

{
  "@context": {
    "@vocab": "https://schema.org/",
    "cr": "https://mlcommons.org/croissant/",
    "ml": "http://ml-schema.org/",
    "sc": "https://schema.org/",
    "dct": "http://purl.org/dc/terms/",
    "data": "https://mlcommons.org/croissant/data/",
    "rai": "https://mlcommons.org/croissant/rai/",
    "format": "https://mlcommons.org/croissant/format/",
    "citeAs": "https://mlcommons.org/croissant/citeAs/",
    "conformsTo": "https://mlcommons.org/croissant/conformsTo/",
    "@language": "en",
    "repeated": "https://mlcommons.org/croissant/repeated/",
    "field": "https://mlcommons.org/croissant/field/",
    "examples": "https://mlcommons.org/croissant/examples/",
    "recordSet": "https://mlcommons.org/croissant/recordSet/",
    "fileObject": "https://mlcommons.org/croissant/fileObject/",
    "fileSet": "https://mlcommons.org/croissant/fileSet/",
    "source": "https://mlcommons.org/croissant/source/",
    "references": "https://mlcommons.org/croissant/references/",
    "key": "https://mlcommons.org/croissant/key/",
    "parentField": "https://mlcommons.org/croissant/parentField/",
    "isLiveDataset": "https://mlcommons.org/croissant/isLiveDataset/",
    "separator": "https://mlcommons.org/croissant/separator/",
    "extract": "https://mlcommons.org/croissant/extract/",
    "subField": "https://mlcommons.org/croissant/subField/",
    "regex": "https://mlcommons.org/croissant/regex/",
    "column": "https://mlcommons.org/croissant/column/",
    "path": "https://mlcommons.org/croissant/path/",
    "fileProperty": "https://mlcommons.org/croissant/fileProperty/",
    "md5": "https://mlcommons.org/croissant/md5/",
    "jsonPath": "https://mlcommons.org/croissant/jsonPath/",
    "transform": "https://mlcommons.org/croissant/transform/",
    "replace": "https://mlcommons.org/croissant/replace/",
    "dataType": "https://mlcommons.org/croissant/dataType/",
    "includes": "https://mlcommons.org/croissant/includes/",
    "excludes": "https://mlcommons.org/croissant/excludes/"
  },
  "@type": "Dataset",
  "name": "Mini immuno dataset",
  "description": "A few samples from the immunology dataset",
  "url": "https://lamin.ai/laminlabs/lamindata/artifact/tCUkRcaEjTjhtozp0000",
  "creator": {
    "@type": "Person",
    "name": "falexwolf"
  },
  "dateCreated": "2025-07-16",
  "cr:projectName": "Mini Immuno Project",
  "datePublished": "2025-07-16",
  "version": "1.0",
  "license": "https://creativecommons.org/licenses/by/4.0/",
  "citation": "Please cite this dataset as: mini immuno (2025)",
  "encodingFormat": "zarr",
  "distribution": [
    {
      "@type": "cr:FileSet",
      "@id": "mini_immuno.anndata.zarr",
      "containedIn": {
        "@id": "directory"
      },
      "encodingFormat": "zarr"
    }
  ],
  "cr:recordSet": [
    {
      "@type": "cr:RecordSet",
      "@id": "#samples",
      "name": "samples",
      "description": "my sample"
    }
  ]
}

dataset1_path

Show code cell output Hide code cell output

PosixPath('mini_immuno.anndata.zarr')

artifact = ln.integrations.curate_from_croissant(croissant_path)

Show code cell output Hide code cell output

! calling anonymously, will miss private instances

! file path mini_immuno.anndata.zarr is not part of a known storage location, will be duplicated to: StorageSettings(root='/home/runner/work/lamin-mlops/lamin-mlops/docs/test-lamin-croissant', uid='5Tqt1lldyO8z')

Project label, license, description, version tag, and file paths are automatically extracted from the Croissant file. More metadata can be supported in the future.

artifact.describe()

Show code cell output Hide code cell output

Artifact: mini_immuno.anndata.zarr (1.0)
|   description: Mini immuno dataset - A few samples from the immunology dataset
├── uid: ljkRXQRE9kX3yKHA0000            run: RXmr7Qc (croissant.ipynb)
│   kind: dataset                        otype: AnnData                
│   hash: N4_3ooEU3qkk97bWijn9XA         size: 21.5 KB                 
│   branch: main                         space: all                    
│   created_at: 2025-11-05 21:34:24 UTC  created_by: anonymous         
│   n_files: 95                                                        
├── storage/path: 
│   /home/runner/work/lamin-mlops/lamin-mlops/docs/test-lamin-croissant/.lamindb/ljkRXQRE9kX3yKHA.anndata.zarr
└── Labels
    └── .projects                       Project                            Mini Immuno Project                     
        .ulabels                        ULabel                             https://creativecommons.org/licenses/by…

ln.finish()

Show code cell output Hide code cell output

→ finished Run('RXmr7QcGEwN73f8V') after 2s at 2025-11-05 21:34:25 UTC