Metadata are written as a python dictionary. They have to meet following requirements:
top level dictionary contains category
,table
and features
category
denotes a group these features belong to for further filtering
table
is a mandatory entry and it denotes to which feature table features from this notebook will be saved to, learn more in
features
is a dictionary containing key
value
pairs where key
is the name of a feature and value is dictionary containing additional metadata
description
- Contains description of given featuretags
- is a list of strings, specifying given featurefillna_with
- Value that nulls are gonna be replaced withmetadata = {
"category": "example",
"table": "table_name"
"features": {
"feature1": {
"description": "Feature 1",
"tags": ["tag1", "tag2"],
"fillna_with": 0
},
"feature2": {
"description": "Feature 2",
"tags": ["tag1", "tag2"]
},
}
}