kozhindev-clearml-wrapper 0.0.2

Wrapper over clearml, developed by ML team of KozhinDev company

KozhinDev ClearML Wrapper

A set of utility functions and decorators for simplifying ClearML integration in your ML projects

Features

• log_metrics – decorator for logging scalar metrics
• get_local_dataset_path – get local path to a dataset file/folder
• prepare_task – initialize or retrieve ClearML tasks (local or remote)
• get_local_model_path – retrieve a local copy of a model artifact

First, install the library: pip install kozhindev_clearml_wrapper

Usage

1. Logging metrics
   Use the @log_metrics decorator to log scalar metrics to ClearML:

from clearml import Task
from kozhindev_clearml_wrapper import log_metrics

task = Task.init(project_name="Demo", task_name="Log Metrics Example")

@log_metrics(task)
def train():
    for i in range(5):
        yield "loss", "train", 0.1 * i, i  # (title, series, metric, iteration)

train()

2. Get Local Dataset Path
   Retrieve a dataset by ID and get the local path to a file or folder inside it:

from kozhindev_clearml_wrapper import get_local_dataset_path

dataset_path = get_local_dataset_path(
    dataset_id="your-dataset-id",
    dataset_name="data.csv"
)

print(dataset_path)

3. Prepare ClearML Task
   Initialize a new task (local) or connect to an existing one (remote):

from kozhindev_clearml_wrapper import prepare_task

# Local task (creates a new one)
task = prepare_task(
    task_type="local",
    project_name="Demo",
    task_name="Local Task Example"
)

# Remote task (connects training parameters to the current ClearML task)
train_params = {"learning_rate": 0.001, "epochs": 10}
task = prepare_task(
    task_type="remote",
    train_params=train_params
)

4. Get Local Model Path
   Download and get the local path to a model artifact from a specific task:

from kozhindev_clearml_wrapper import get_local_model_path

path_to_model = get_local_model_path(
    task_id="your-task-id",
    artifact_name="trained_models"
)

print(path_to_model)