cogflow-lite 0.1.4

Lighter version of COG modules

cogflow_lite

Cogflow_lite module sets up a pipeline for handling datasets and machine learning models using multiple plugins. It includes functions for creating, registering, evaluating, and serving models, as well as managing datasets.

Key components include:

Mlflow Plugin: For model tracking, logging, and evaluation. Kubeflow Plugin: For pipeline management and serving models. Dataset Plugin: For dataset registration and management. Model Plugin: For saving model details. Configurations: Constants for configuration like tracking URIs, database credentials, etc.

Key Functions:

Model Management

register_model: Register a new model. log_model: Log a model. load_model: Load a model. delete_registered_model: Delete a registered model. create_registered_model: Create a new registered model. create_model_version: Create a new version of a registered model.

Run Management

start_run: Start a new. end_run: End the current. log_param: Log a parameter to the current run. log_metric: Log a metric to the current run.

Evaluation and Autologging

evaluate: Evaluate a model. autolog: Enable automatic logging of parameters, metrics, and models.

Search and Query

search_registered_models: Search for registered models. search_model_versions: Search for model versions. get_model_latest_version: Get the latest version of a registered model. get_artifact_uri: Get the artifact URI of the current or specified run.

Dataset Management

link_model_to_dataset: Link a model to a dataset. save_dataset_details: Save dataset details. save_model_details_to_db: Save model details to the database.

Pipeline and Component Management

pipeline: Create a new Kubeflow pipeline. create_component_from_func: Create a Kubeflow component from a function. client: Get the Kubeflow client. load_component_from_url: Load a Kubeflow component from a URL.

Model Serving

serve_model_v1: Serve a model using Kubeflow V1. serve_model_v2: Serve a model using Kubeflow V2. get_model_url: Get the URL of a served model. delete_served_model: Delete a served model.

MinIO Operations

create_minio_client: Create a MinIO client. query_endpoint_and_download_file: Query an endpoint and download a file from MinIO. save_to_minio: Save file content to MinIO. delete_from_minio: Delete an object from MinIO.

Dataset Registration

register_dataset: Register a dataset.

Getting Started

To begin, import cogflow from the CogFlow module:

import cogflow_lite

Explore the Capabilities of cogflow_lite

• List Attributes and Methods: Understand the cogflow_lite module better with:

  print(dir(cogflow_lite))
  

• Get Documentation: For a comprehensive guide on the cogflow_lite, use:

  help(cogflow_lite)
  

Environment Variables

To maximize the functionality of CogFlow_lite, set the following environment variables:

• Mlflow Configuration:

  • MLFLOW_TRACKING_URI: The URI of the Mlflow tracking server.
  • MLFLOW_S3_ENDPOINT_URL: The endpoint URL for the AWS S3 service.
  • ACCESS_KEY_ID: The access key ID for AWS S3 authentication.
  • SECRET_ACCESS_KEY: The secret access key for AWS S3 authentication.

• Machine Learning Database:

  • ML_DB_USERNAME: Username for connecting to the machine learning database.
  • ML_DB_PASSWORD: Password for connecting to the machine learning database.
  • ML_DB_HOST: Host address for the machine learning database.
  • ML_DB_PORT: Port number for the machine learning database.
  • ML_DB_NAME: Name of the machine learning database.

• MinIO Configuration:

  • MINIO_ENDPOINT_URL: The endpoint URL for the MinIO service.
  • MINIO_ACCESS_KEY: The access key for MinIO authentication.
  • MINIO_SECRET_ACCESS_KEY: The secret access key for MinIO authentication.

---

By setting the environment variables correctly, you can fully utilize the features and functionalities of the CogFlow framework for your cognitive and machine learning tasks.