Skip to main content

Documentation Index

Fetch the complete documentation index at: https://arize-ax.mintlify.dev/docs/llms.txt

Use this file to discover all available pages before exploring further.

View Source on Github Arize class to organize and map column names containing model data within your Pandas dataframe to Arize. Import and initialize Arize Schema from arize.utils.types 
from arize.utils.types import Schema

class Schema(
    prediction_id_column_name: Optional[str] = None
    feature_column_names: Optional[Union[List[str], TypedColumns]] = None
    tag_column_names: Optional[Union[List[str], TypedColumns]] = None
    timestamp_column_name: Optional[str] = None
    prediction_label_column_name: Optional[str] = None
    prediction_score_column_name: Optional[str] = None
    actual_label_column_name: Optional[str] = None
    actual_score_column_name: Optional[str] = None
    shap_values_column_names: Optional[Dict[str, str]] = None
    embedding_feature_column_names: Optional[Dict[str, EmbeddingColumnNames]] = None  # type:ignore
    prediction_group_id_column_name: Optional[str] = None
    rank_column_name: Optional[str] = None
    attributions_column_name: Optional[str] = None
    relevance_score_column_name: Optional[str] = None
    relevance_labels_column_name: Optional[str] = None
    object_detection_prediction_column_names: Optional[ObjectDetectionColumnNames] = None
    object_detection_actual_column_names: Optional[ObjectDetectionColumnNames] = None
    prompt_column_names: Optional[Union[str, EmbeddingColumnNames]] = None
    response_column_names: Optional[Union[str, EmbeddingColumnNames]] = None
    prompt_template_column_names: Optional[PromptTemplateColumnNames] = None
    llm_config_column_names: Optional[LLMConfigColumnNames] = None
    llm_run_metadata_column_names: Optional[LLMRunMetadataColumnNames] = None
    retrieved_document_ids_column_name: Optional[List[str]] = None
    multi_class_threshold_scores_column_name: Optional[str] = None
)
ParameterData TypeExpected Type In ColumnDescription
prediction_id_column_namestrContents must be a string limited to 128 characters(Optional) A unique string to identify a prediction event. Required to match a prediction to delayed actuals or feature importances in Arize. If the column is not provided, Arize will generate a random prediction id.
feature_column_namesList[str] or TypedColumnsFeature values can be int, float, string, list of strings(Optional) Column names for features. If TypedColumns is used, the columns will be cast to the provided types prior to logging.
embedding_feature_column_namesDict[str, EmbeddingColumnNames]Learn more here(Optional) Dictionary mapping embedding display names to EmbeddingColumnNames objects
timestamp_column_namestrThe content of this column must be int Unix Timestamps in seconds(Optional) Column name for timestamps
prediction_label_column_namestrThe content of this column must be convertible to string(Optional) Column name for categorical prediction values
prediction_score_column_namestrThe content of this column must be int/float.For Multi-Class models, content of this column must be a dictionary, mapping class name to int/float prediction scores.(Optional Column name for numeric prediction values
actual_label_column_namestrThe content of this column must be convertible to string(Optional) Column name for categorical ground truth values
actual_score_column_namestrThe content of this column must be int/float.For Multi-Class models, content of this column must be a dictionary, mapping class name to int/float actual scores.(Optional) Column name for numeric ground truth
tag_column_namesList[str] or TypedColumnsTag values can be int, float, string. LImited to 1k values(Optional) Column names for tags. If TypedColumns is used, the columns will be cast to the provided types prior to logging.
shap_values_column_namesDict[str,str]The content of this column must be int/float(Optional) dict of k-v pairs where k is the feature_colname and v is the corresponding shap_val_col_name. For example, your dataframe contains features columnsfeat1, feat2, feat3,...and corresponding shap value columns feat1_shap, feat2_shap, feat3_shap,... You want to set shap_values_column_names = {"feat1": "feat1shap", "feat2": "feat2_shap:", "feat3": "feat3_shap"}
prediction_group_id_column_namestrThe content of this column must be string and is limited to 128 characters(Required*) Column name for ranking groups or lists in ranking models *for ranking models only
rank_column_namestrThe content of this column must be integer between 1-100(Required*) Column name for rank of each element on the its group or list *for ranking models only
relevance_score_column_namestrThe content of this column must be int/float(Required*) Column name for ranking model type numeric ground truth values *for ranking models only
relevance_labels_column_namestrThe content of this column must be a string(Required*) Column name for ranking model type categorical ground truth values *for ranking models only
object_detection_prediction_column_namesObjectDetectionColumnNamesLearn more hereObjectDetectionColumnNames object containing information defining the predicted bounding boxes’ coordinates, categories, and scores.
object_detection_actual_column_namesObjectDetectionColumnNamesLearn more hereObjectDetectionColumnNames object containing information defining the actula bounding boxes’ coordinates, categories, and scores.
prompt_column_namesEmbeddingColumnNamesLearn more hereEmbeddingColumnNames object containing the embedding vector data (required) and raw text (optional) for the input text your model acts on
response_column_namesEmbeddingColumnNamesLearn more hereEmbeddingColumnNames object containing the embedding vector data (required) and raw text (optional) for the text your model generates
prompt_template_column_namesPromptTemplateColumnNamesLearn more herePromptTemplateColumnNames object containing the prompt template and prompt template version, both optional
llm_config_column_namesLLMConfigColumnNamesLearn more hereLLMConfigColumnNames object containing the LLM model name (optional) and its hyper-parameters (optional) used at inference time
llm_run_metadata_column_namesLLMRunMetadataColumnNamesLearn more hereLLMRunMetadata object containing metadata about the LLM inference, i.e., token counts and response latency
retrieved_document_ids_column_namestrThe contents of this column must be list of entries convertible to stringsColumn name for retrieved document ids
multi_class_threshold_scores_column_namestrContents of this column must be a dictionary mapping string class names to float scores.Learn more here(Optional) Column name used only for Multi-Label Multi-Class models and determines the minimum prediction value for a class to be considered a positive prediction.

​
Code Example

prediction idfeature_1feature_2tag_1tag_2prediction_tsprediction_labelactual_labelembedding
1fcd50f4689ca[ca, ak]female251637538845No ClaimsNo Claims[1.27346, -0.2138, …]
schema = Schema(
    prediction_id_column_name="prediction id",
    feature_column_names=["feature_1", "feature_2"],
    tag_column_names=TypedColumns(
        to_str=["tag_1", "tag_2"],
    ),
    timestamp_column_name="prediction_ts",
    prediction_label_column_name="prediction_label",
    prediction_score_column_name="prediction_score",
    actual_label_column_name="actual_label",
    actual_score_column_name="actual_score",
    shap_values_column_names=shap_values_column_names=dict(zip("feature_1", shap_cols)),
    embedding_feature_column_names=EmbeddingColumnNames(
        vector_column_name="embedding",
    ),
    prediction_group_id_column_name="group_example_name",
    rank_column_name="example_rank",
    relevance_score_column_name="relevance_score",
    relevance_labels_column_name="actual_relevancy",
)