Skip to main content

Creating your first Trial

Learn how to programmatically create and run a trial using the EvoML Client.

Step 1: Connecting into the platform​

Use your credentials to connect to the EvoML platform.

from typing import Final
import evoml_client as ec
import pandas as pd

# Pease replace with your deploy-platform URL
API_URL: Final[str] = "https://evoml.ai"

# Please replace with your username
USERNAME: Final[str] = ""
# Please replace with your password
PASSWORD: Final[str] = ""

# Connect to the platform
ec.init(base_url=API_URL, username=USERNAME, password=PASSWORD)

Step 2: Upload a Dataset​

Upload the heart.csv dataset to the EvoML platform.

dataset = ec.Dataset.from_file("./data/heart.csv")
dataset.put()

dataset_id = dataset.dataset_id

print(f"Dataset ID: {dataset_id}")

Step 3: create the configuration of the Trial​

Define the trial parameters: task type, budget, objective, and dataset ID.

from evoml_client.trial_conf_models import BudgetMode

config = ec.TrialConfig.with_default(
    task=ec.MlTask.classification.value,
    budget_mode=BudgetMode.fast,
    loss_funcs=["F1"],
    dataset_id=dataset_id,
)

Set up the trial with the target column and trial name.

trial, dataset = ec.Trial.from_dataset_id(
    dataset_id,
    target_col="Sex",
    trial_name="Classification - Sex",
    config=config,
)

Run the trial in blocking mode, waiting for execution to complete.

trial.run(timeout=900)

Retrieve and preview the models generated by the trial.

models = trial.get_models()
print(models)