cmflib.cmf.Cmf.DataSlice

A data slice represents a named subset of data. It can be used to track performance of an ML model on different slices of the training or testing dataset splits. This can be useful from different perspectives, for instance, to mitigate model bias.

Instances of data slices are not meant to be created manually by users. Instead, use Cmf.create_dataslice method.

Source code in cmflib/cmf.py

def __init__(self, name: str, writer):
    self.props = {}
    self.name = name
    self.writer = writer

add_data(path, custom_properties=None)

Add data to create the dataslice. Currently supported only for file abstractions. Pre-condition - the parent folder, containing the file should already be versioned. Example:

#dataslice.add_data(f"data/raw_data/{j}.xml)

Args: path: Name to identify the file to be added to the dataslice. custom_properties: Properties associated with this datum.

Source code in cmflib/cmf.py

def add_data(
    self, path: str, custom_properties: t.Optional[t.Dict] = None
) -> None:
    """Add data to create the dataslice.
    Currently supported only for file abstractions. Pre-condition - the parent folder, containing the file
        should already be versioned.
    Example:
        ```python
        #dataslice.add_data(f"data/raw_data/{j}.xml)
        ```
    Args:
        path: Name to identify the file to be added to the dataslice.
        custom_properties: Properties associated with this datum.
    """

    self.props[path] = {}
    self.props[path]['hash'] = dvc_get_hash(path)
    parent_path = path.rsplit("/", 1)[0]
    self.data_parent = parent_path.rsplit("/", 1)[1]
    if custom_properties:
        for k, v in custom_properties.items():
            self.props[path][k] = v

commit(custom_properties=None)

Commit the dataslice. The created dataslice is versioned and added to underneath data versioning software. Example:

dataslice.commit()
```

Args: custom_properties: Dictionary to store key value pairs associated with Dataslice Example{"mean":2.5, "median":2.6}

Source code in cmflib/cmf.py

def commit(self, custom_properties: t.Optional[t.Dict] = None) -> None:
    """Commit the dataslice.
    The created dataslice is versioned and added to underneath data versioning software.
    Example:

        dataslice.commit()
        ```
    Args:
        custom_properties: Dictionary to store key value pairs associated with Dataslice
        Example{"mean":2.5, "median":2.6}
    """

    logging_dir = change_dir(self.writer.cmf_init_path)
    # code for nano cmf
    # Assigning current file name as stage and execution name
    current_script = sys.argv[0]
    file_name = os.path.basename(current_script)
    name_without_extension = os.path.splitext(file_name)[0]
    # create context if not already created
    if not self.writer.child_context:
        self.writer.create_context(pipeline_stage=name_without_extension)
        assert self.writer.child_context is not None, f"Failed to create context for {self.pipeline_name}!!"

    # create execution if not already created
    if not self.writer.execution:
        self.writer.create_execution(execution_type=name_without_extension)
        assert self.writer.execution is not None, f"Failed to create execution for {self.pipeline_name}!!"

    directory_path = os.path.join(self.writer.ARTIFACTS_PATH, self.writer.execution.properties["Execution_uuid"].string_value.split(',')[0], self.writer.DATASLICE_PATH)
    os.makedirs(directory_path, exist_ok=True)
    custom_props = {} if custom_properties is None else custom_properties
    git_repo = git_get_repo()
    dataslice_df = pd.DataFrame.from_dict(self.props, orient="index")
    dataslice_df.index.names = ["Path"]
    dataslice_path = os.path.join(directory_path,self.name)
    dataslice_df.to_parquet(dataslice_path)
    existing_artifact = []

    commit_output(dataslice_path, self.writer.execution.id)
    c_hash = dvc_get_hash(dataslice_path)
    if c_hash == "":
        print("Error in getting the dvc hash,return without logging")
        return

    dataslice_commit = c_hash
    url = dvc_get_url(dataslice_path)
    dvc_url_with_pipeline = f"{self.writer.parent_context.name}:{url}"
    if c_hash and c_hash.strip():
        existing_artifact.extend(
            self.writer.store.get_artifacts_by_uri(c_hash))
    if existing_artifact and len(existing_artifact) != 0:
        print("Adding to existing data slice")
        # Haven't added event type in this if cond, is it not needed??
        slice = link_execution_to_input_artifact(
            store=self.writer.store,
            execution_id=self.writer.execution.id,
            uri=c_hash,
            input_name=dataslice_path + ":" + c_hash,
        )
    else:
        slice = create_new_artifact_event_and_attribution(
            store=self.writer.store,
            execution_id=self.writer.execution.id,
            context_id=self.writer.child_context.id,
            uri=c_hash,
            name=dataslice_path + ":" + c_hash,
            type_name="Dataslice",
            event_type=mlpb.Event.Type.OUTPUT,
            properties={
                "git_repo": str(git_repo),
                # passing c_hash value to commit
                "Commit": str(dataslice_commit),
                "url": str(dvc_url_with_pipeline),
            },
            artifact_type_properties={
                "git_repo": mlpb.STRING,
                "Commit": mlpb.STRING,
                "url": mlpb.STRING,
            },
            custom_properties=custom_props,
            milliseconds_since_epoch=int(time.time() * 1000),
        )

    custom_props["git_repo"] = git_repo
    custom_props["Commit"] = dataslice_commit
    self.writer.execution_label_props["git_repo"] = git_repo
    self.writer.execution_label_props["Commit"] = dataslice_commit
    if self.writer.graph:
        self.writer.driver.create_dataslice_node(
            self.name, dataslice_path + ":" + c_hash, c_hash, self.data_parent, custom_props
        )
    os.chdir(logging_dir)
    return slice