mflux-ai

This is the official mflux-ai python library for MFlux.ai

Features

• Fetch connection strings and tell MLflow how to connect with MFlux.ai
• Download and upload objects/datasets from/to the MFlux.ai cloud service

Quickstart

Installation

pip install mflux-ai

Basic usage

import mflux_ai

mflux_ai.init("INSERT_YOUR_PROJECT_TOKEN_HERE")

# MLflow now knows how to connect with your project server, hosted on MFlux.ai

Store and retrieve datasets

my_dataset = np.zeros(shape=(10000, 100), dtype=np.float32)
dataset_filename = "my-dataset.pkl"

mflux_ai.put_dataset(my_dataset, dataset_filename)

my_loaded_dataset = mflux_ai.get_dataset(dataset_filename)

assert_array_equal(my_dataset, my_loaded_dataset)

Fetch the best run in a model group defined in MFlux.ai. Runs are evaluated on one or two metrics defined in MFlux.ai.

best_run = mflux_ai.get_best_run("classification")
assert isinstance(run, mlflow.entities.run.Run)

# You can now use the run class to fetch info about the run, for instance metrics and run id
metrics = best_run.data.metrics
run_id = best_run.info.run_uuid

# You can fetch the artifact_uri to load the model
model = mlflow.sklearn.load_model(best_run.info.artifact_uri + "/model")