Skip to main content
Version: 25.9

snorkelai.sdk.develop.Slice

final class snorkelai.sdk.develop.Slice(dataset, slice_uid, name, description=None, templates=None, graph=None)

Bases: Base

Represents a slice within a Snorkel dataset for identifying and managing subsets of datapoints.

A slice is a logical subset of datapoints within a dataset that can be created either manually by adding specific datapoints, or programmatically using slicing functions defined through templates and configurations. Slices are essential for data analysis, model evaluation, and targeted data operations within Snorkel workflows.

Key capabilities:

  • Manual datapoint management through add/remove operations

  • Programmatic datapoint identification using configurable slicing functions

This class provides methods for creating, retrieving, updating, and managing slice membership. Slice objects should not be instantiated directly - use the create() or get() class methods instead.

Read more in Using data slices.

Using the Slice class requires the following import:

from snorkelai.sdk.develop import Slice

__init__​

__init__(dataset, slice_uid, name, description=None, templates=None, graph=None)

Create a Slice object in-memory with necessary properties. This constructor should not be called directly, and should instead be accessed through the create() and get() methods

Parameters​

NameTypeDefaultInfo
datasetUnion[str, int]The UID or name for the dataset within Snorkel.
slice_uidintThe UID for the slice within Snorkel.
namestrThe name of the slice.
descriptionOptional[str]NoneThe description of the slice.
templatesOptional[List[Dict[str, Any]]]NoneConfiguration defining slicing functions for programmatic datapoint identification.
graphOptional[List[Any]]NoneA representation of the slicing function’s structure.

Methods

__init__(dataset, slice_uid, name[, ...])Create a Slice object in-memory with necessary properties.
add_x_uids(x_uids)Adds datapoints to a slice.
create(dataset, name[, description, ...])Creates a slice for a dataset.
delete(dataset, slice_uid)Deletes a slice from a dataset.
get(dataset, slice)Retrieves a slice by UID or name.
get_x_uids()Retrieves the UIDs of the datapoints in the slice.
list(dataset)Retrieves all slices for a dataset.
remove_x_uids(x_uids)Removes datapoints from a slice.
update([name, description, templates, graph])Updates the slice properties.

Attributes

dataset_uidReturn the UID of the dataset that the slice belongs to
descriptionReturn the description of the slice
nameReturn the name of the slice
slice_uidReturn the UID of the slice
uidReturn the UID of the slice

add_x_uids​

add_x_uids(x_uids)

Adds datapoints to a slice.

Parameters​

NameTypeDefaultInfo
x_uidsList[str]List of UIDs of the datapoints you want to add to the slice.

Raises​

Exception – If other server errors occur during the operation.

Return type​

None

Example​

from snorkelai.sdk.develop import Slice
slice = Slice.get(dataset=1, slice=20)
slice.add_x_uids(["uid1", "uid2", "uid3"])

create​

classmethod create(dataset, name, description='', templates=None, graph=None)

Creates a slice for a dataset.

Slices are used to identify a subset of datapoints in a dataset. You can add datapoints to a slice manually, or if you define a config, you can add datapoints programmatically. Slice membership can contain both manual and programmatic identified datapoints.

Parameters​

NameTypeDefaultInfo
datasetUnion[str, int]The UID or name for the dataset within Snorkel.
namestrThe name of the slice.
descriptionstr''A description of the slice, by default the empty string.
templatesOptional[List[Dict[str, Any]]]NoneA list of template dictionaries, by default None, you can reference the schema in the template module for constructing these templates. These templates are used to define the Slicing Function for the slice, allowing it to programmatically add datapoints to the slice membership.
graphOptional[List[Any]]NoneA representation of the slicing function’s structure, by default None.

Returns​

The slice object.

Return type​

Slice

Raises​

  • ValueError – If the dataset doesn’t exist or cannot be found by name/UID.

  • ValueError – If the slice name is a reserved name or already exists for the dataset.

  • ValueError – If there are other validation or server errors during slice creation.

Examples​

Example 1​

Create a simple slice without templates:

from snorkelai.sdk.develop import Slice
slice = Slice.create(
    dataset=1,
    name="my_slice",
    description="A slice for testing purposes"
)

Example 2​

Create a slice from a keyword template. This example creates a slice containing all datapoints with the string capital in the Instruction field:

from snorkelai.sdk.develop import Slice
from snorkelai.sdk.utils.graph import DEFAULT_GRAPH
from snorkelai.templates.keyword_template import KeywordTemplateSchema

keyword_template = KeywordTemplateSchema(
    operator="CONTAINS",
    keywords=["capital"],
    field="Instruction",
    case_sensitive=False,
    tokenize=True,
)
template_config_dict = {
    **keyword_template.to_dict(),
    "template_type": "keyword"
}
slice = Slice.create(
    dataset=1,
    name="slice_name",
    description="description",
    templates=[
        {
            "transform_type": "dataset_template_filter",
            "config": {
                "transform_config_type": "template_filter_schema",
                "filter_type": "text_template",
                "filter_config_type": "dataset_text_template",
                "dataset_uid": 1,
                "template_config": template_config_dict,
             },
        },
        ],
    graph=DEFAULT_GRAPH,
)

delete​

classmethod delete(dataset, slice_uid)

Deletes a slice from a dataset.

Parameters​

NameTypeDefaultInfo
datasetUnion[str, int]The UID or name for the dataset within Snorkel.
slice_uidintThe UID of the slice to delete.

Raises​

  • ValueError – If the dataset or slice doesn’t exist or cannot be found by UID.

  • ValueError – If there are other validation or server errors during slice deletion.

Return type​

None

Example​

from snorkelai.sdk.develop import Slice
Slice.delete(dataset=1, slice_uid=20)

get​

classmethod get(dataset, slice)

Retrieves a slice by UID or name.

Parameters​

NameTypeDefaultInfo
datasetUnion[str, int]The UID or name for the dataset within Snorkel.
sliceUnion[str, int]The UID or name of the slice.

Returns​

The slice object.

Return type​

Slice

Raises​

ValueError – If no slice is found with the given UID or name.

Example​

from snorkelai.sdk.develop import Slice
slice = Slice.get(dataset=1, slice=20)

get_x_uids​

get_x_uids()

Retrieves the UIDs of the datapoints in the slice.

Returns​

List of UIDs of the datapoints in the slice.

Return type​

List[str]

Raises​

Exception – If other server errors occur during the operation.

Example​

from snorkelai.sdk.develop import Slice
slice = Slice.get(dataset=1, slice=20)
x_uids = slice.get_x_uids()

list​

classmethod list(dataset)

Retrieves all slices for a dataset.

Parameters​

NameTypeDefaultInfo
datasetUnion[str, int]The UID or name for the dataset within Snorkel.

Returns​

A list of all the slices available for that dataset.

Return type​

List[Slice]

Raises​

ValueError – If no dataset is found with the given id.

Example​

from snorkelai.sdk.develop import Slice
slices = Slice.list(dataset=1)

remove_x_uids​

remove_x_uids(x_uids)

Removes datapoints from a slice.

Parameters​

NameTypeDefaultInfo
x_uidsList[str]List of UIDs of the datapoints you want to remove from the slice.

Raises​

Exception – If other server errors occur during the operation.

Return type​

None

Example​

from snorkelai.sdk.develop import Slice
slice = Slice.get(dataset=1, slice=20)
slice.remove_x_uids(["uid1", "uid2", "uid3"])

update​

update(name=None, description=None, templates=None, graph=None)

Updates the slice properties.

Parameters​

NameTypeDefaultInfo
nameOptional[str]NoneThe new name for the slice, by default None.
descriptionOptional[str]NoneThe new description for the slice, by default None.
templatesOptional[List[Dict[str, Any]]]NoneA list of template dictionaries for the slice, by default None.
graphOptional[List[str]]NoneA list of strings representing the graph structure for the slice, by default None.

Raises​

ValueError – If there are other errors during slice update.

Return type​

None

Example​

from snorkelai.sdk.develop import Slice
slice = Slice.get(dataset=1, slice=20)
slice.update(name="new_name", description="updated description")
property dataset_uid: int

Return the UID of the dataset that the slice belongs to

property description: str | None

Return the description of the slice

property name: str

Return the name of the slice

property slice_uid: int

Return the UID of the slice

property uid: int

Return the UID of the slice