Skip to content

ApiKeyExpirationTime

ApiKeyExpirationTime()

Fields:

  • ONE_MONTH
  • THREE_MONTHS
  • SIX_MONTHS
  • ONE_YEAR
  • NEVER

BaseML3Enum

BaseML3Enum()

Base class for all enums in the ML3 Platform SDK

BooleanLicenceFeature

BooleanLicenceFeature()

Boolean licence feature

Fields: - EXPLAINABILITY Whether the company has access to explainability reports - MONITORING Whether the company has monitoring feature enabled - MONITORING_METRICS Whether the company has monitoring metrics feature enabled - SEGMENTED_MONITORING Whether the company has segmented monitoring feature enabled - RETRAINING Whether the company has retraining feature enabled - TOPIC_ANALYSIS Whether the company has topic analysis feature enabled - RAG_EVALUATION Whether the company has RAG evaluation feature enabled - LLM_SECURITY Whether the company has LLM security feature enabled - BUSINESS Whether the company has business feature enabled

ColumnRole

ColumnRole()

Column role enum Describe the role of a column

Fields: - INPUT - INPUT_MASK - METADATA - PREDICTION - TARGET - ERROR - ID - TIME_ID - INPUT_ADDITIONAL_EMBEDDING - TARGET_ADDITIONAL_EMBEDDING - PREDICTION_ADDITIONAL_EMBEDDING - USER_INPUT - RETRIEVED_CONTEXT

ColumnSubRole

ColumnSubRole()

This enum describes the subrole of a column in the data schema For instance, it's used in RAG tasks to distinguish between user input and retrieved context

Subroles for ColumnRole.INPUT in RAG settings: - RAG_USER_INPUT - RAG_RETRIEVED_CONTEXT - RAG_SYS_PROMPT

Subroles for ColumnRole.INPUT in TIMESERIES settings: - SEASONALITY - TREND - REGRESSOR

Subroles for ColumnRole.PREDICTION: - MODEL_PROBABILITY - OBJECT_LABEL_PREDICTION - OBJECT_TEXT_PREDICTION

Subroles for ColumnRole.TARGET: - OBJECT_LABEL_TARGET - OBJECT_TEXT_TARGET

Subroles for ColumnRole.METADATA in RAG settings: - RAG_SESSION_ID - RAG_TURN_ID

Currency

Currency()

Currency of to use for the Task

Fields: - EURO - DOLLAR

DataBatchType

DataBatchType()

Defines the type of the uploaded data batch.

Fields:

TRAINING VALIDATION TEST PRODUCTION

DataStructure

DataStructure()

Represents the typology of the data to send

Fields:

  • TABULAR
  • IMAGE
  • TEXT
  • EMBEDDING

DataType

DataType()

Data type enum Describe data type of input

Fields: - FLOAT - STRING - CATEGORICAL - INTEGER - ARRAY_1 - ARRAY_2 - ARRAY_3

DetectionEventActionType

DetectionEventActionType()

Fields:

  • DISCORD_NOTIFICATION
  • SLACK_NOTIFICATION
  • EMAIL_NOTIFICATION
  • TEAMS_NOTIFICATION
  • MQTT_NOTIFICATION
  • RETRAIN
  • NEW_PLOT_CONFIGURATION
  • AWS_EVENT_BRIDGE_NOTIFICATION
  • GCP_PUBSUB_NOTIFICATION
  • AZURE_EVENT_GRID_NOTIFICATION

DetectionEventSeverity

DetectionEventSeverity()

Fields:

  • LOW
  • MEDIUM
  • HIGH

DetectionEventType

DetectionEventType()

Fields:

  • WARNING_OFF
  • WARNING_ON
  • DRIFT_ON
  • DRIFT_OFF

ExternalIntegration

ExternalIntegration()

An integration with a 3rd party service provider

Fields: - AWS - GCP - AZURE - AWS_COMPATIBLE - GOOGLE_GENAI - GOOGLE_VERTEXAI - OPENAI - AZURE_OPENAI - ANTHROPIC

FileType

FileType()

Fields:

  • CSV
  • JSON
  • PARQUET
  • PNG
  • JPG
  • NPY

FolderType

FolderType()

Type of folder

Fields

  • UNCOMPRESSED
  • TAR
  • ZIP

GenericMonitoringDimension

GenericMonitoringDimension()

Generic monitoring dimensions that wraps all the other signals that can be monitored

ImageMode

ImageMode()

Image mode enumeration

Fields: - RGB - RGBA - GRAYSCALE

JobStatus

JobStatus()

Enum containing all the job's status that a client can see

Fields:

  • IDLE
  • STARTING
  • RUNNING
  • COMPLETED
  • ERROR

KPIStatus

KPIStatus()

Fields:

  • NOT_INITIALIZED
  • OK
  • WARNING
  • DRIFT

LLMProvider

LLMProvider()

Enumerator of the providers of LLMs we can use to generate text both for RAG and LLM.

ModelMetricName

ModelMetricName()

Name of the model metrics that is associated with the model

Fields: - RMSE - RSQUARE - ACCURACY - AVERAGE_PRECISION

MonitoringEvaluationMetric

MonitoringEvaluationMetric()

Metric computed at batch level and without referring to a task.

Differently from MonitoringMetric, that are extractions from quantities (MonitoringTarget) for each sample, the MonitoringEvaluationMetrics are computed for a set of data and not for the single sample.

Each task type has a list of metrics.

BINARY_CLASSIFICATION, MULTICLASS_CLASSIFICATION, MULTILABEL_CLASSIFICATION

  • Accuracy
  • Precision
  • Recall
  • F1-score
  • ROC-AUC (requires model probability)
  • PR-AUC (requires model probability)
  • LogLoss (requires model probability)
  • Balanced Accuracy

REGRESSION

  • MAE
  • RMSE
  • R square

CLUSTERING

  • Silhouette score
  • Calinski-Harabasz Index
  • Adjusted Rand Index (requires ground truth)
  • NMI / V-measure (requires ground truth)

ANOMALY DETECTION

  • ROC-AUC (requires model probability)
  • PR-AUC (requires model probability)
  • FPR @ TPR (requires model probability)

TIMESERIES

  • MAE
  • RMSE
  • MAPE

MonitoringMetric

MonitoringMetric()

Tabular: - FEATURE

Text: - TEXT_TOXICITY - TEXT_EMOTION - TEXT_SENTIMENT - TEXT_LENGTH

Model probabilistic output: - MODEL_PERPLEXITY - MODEL_ENTROPY - MODEL_IMAGE_ENTROPY

Error: - LOG_LIKELIHOOD: likelihood of target sample for distribution induced by the model

Image: - IMAGE_BRIGHTNESS - IMAGE_CONTRAST - IMAGE_FOCUS - IMAGE_BLUR - IMAGE_COLOR_VARIATION - IMAGE_COLOR_CONTRAST

Object detection and semantic segmentation: (position wrt Cartesian axis with origin in the center of the image)

MonitoringStatus

MonitoringStatus()

Fields:

  • OK
  • WARNING
  • DRIFT

MonitoringTarget

MonitoringTarget()

Fields:

  • ERROR
  • INPUT
  • CONCEPT
  • PREDICTION
  • INPUT_PREDICTION
  • USER_INPUT
  • RETRIEVED_CONTEXT
  • USER_INPUT_RETRIEVED_CONTEXT
  • USER_INPUT_MODEL_OUTPUT
  • MODEL_OUTPUT_RETRIEVED_CONTEXT
  • CHARACTER_ERROR_RATE
  • WORD_ERROR_RATE

NumericLicenceFeature

NumericLicenceFeature()

Numeric licence feature

Fields: - MAX_TASKS Maximum number of tasks that the company can have - MAX_USERS Maximum number of users that the company can have - DAILY_DATA_BATCH_UPLOAD Maximum number of data batches that the company can upload in a day. Only considers production data batches.

OcrMode

OcrMode()

Ocr mode enumeration

Fields: - PLAIN_TEXT - WITH_LABELS

ProductKeyStatus

ProductKeyStatus()

Status of a product key

Fields:: - NEW = generated but not yet used product key - VALIDATING = validation requested from client - IN_USE = validated product key, client activated

RetrainTriggerType

RetrainTriggerType()

Enumeration of the possible retrain triggers

Fields:: - AWS_EVENT_BRIDGE - GCP_PUBSUB - AZURE_EVENT_GRID

SegmentOperator

SegmentOperator()

Segment operator for segmentation rules. Fields: - IN: the given rule is verified if the field is in the list of values - OUT: the given rule is verified if the field is not in the list of values

SemanticSegTargetType

SemanticSegTargetType()

Format of the target and prediction for the semantic segmentation task.

POLYGON: each identified object is represented by the vertices of the polygon

StoragePolicy

StoragePolicy()

Enumeration that specifies the storage policy for the data sent to ML cube Platform

Fields: cloud it needs to read data

StoringDataType

StoringDataType()

Fields:

  • HISTORICAL
  • REFERENCE
  • PRODUCTION
  • KPI

SubscriptionType

SubscriptionType()

Type of subscription plan of a company

Fields:: - CLOUD: subscription plan for web app or sdk access - EDGE: subscription plan for edge deployment

SuggestionType

SuggestionType()

Enum to specify the preferred type of suggestion

Fields: - SAMPLE_WEIGHTS - RESAMPLED_DATASET

TaskType

TaskType()

Fields:

  • REGRESSION
  • CLASSIFICATION_BINARY
  • CLASSIFICATION_MULTICLASS
  • CLASSIFICATION_MULTILABEL
  • RAG
  • OBJECT_DETECTION
  • SEMANTIC_SEGMENTATION
  • CLUSTERING

TextLanguage

TextLanguage()

Enumeration of text language used in nlp tasks.

Fields: - ITALIAN - ENGLISH - MULTILANGUAGE

ThresholdValueType

ThresholdValueType()

Supported threshold input types for monitoring thresholds.

Fields: - INT An integer threshold - FLOAT A floating point threshold - PERCENTAGE A percentage threshold (0-100)

TimeseriesMode

TimeseriesMode()

Define how regressors, seasonality or trend are used in the timeseries.

UserCompanyRole

UserCompanyRole()

Fields:

  • COMPANY_OWNER
  • COMPANY_ADMIN
  • COMPANY_USER
  • COMPANY_NONE

UserProjectRole

UserProjectRole()

Fields:

  • PROJECT_ADMIN
  • PROJECT_USER
  • PROJECT_VIEW