python-mlboardclient

Python lib for interacting ml-board component.

Installation

From PyPi

pip install python-mlboardclient

From github (fresh master branch)

pip install 'git+https://github.com/kuberlab/python-mlboardclient.git'

Usage

from mlboardclient.api import client

# Default url is http://mlboard-v2.kuberlab:8082/api/v2
# No need to pass any url if instantiate client from ml-project (Jupyter/Task)
ml = client.Client()
apps = ml.apps.list()
[<mlboardclient.api.v2.apps.App object at 0x7f0b554b5f90>]

app = apps[0]
# Get tasks from config
app.tasks
[<Task name=model build=None status=undefined>]

task = app.tasks[0]

# Run & wait task
task.run()
<Task name=model build=4 status=Succeeded>

# Get tasks from API
app.get_tasks()
[<Task name=model build=1 status=Failed>, <Task name=model build=2 status=Failed>,
<Task name=model build=3 status=Failed>, <Task name=model build=4 status=Succeeded>]

Model upload

    ml.model_upload('my-model', '1.0.0', '/model/dir')
    # If the model uploading is not executing in scope of project task,
    # need to specify workspace_name and project_name explicitly:
    ml.model_upload(
        'my-model',
        '1.0.0',
        '/model/dir',
        workspace_name='demo',
        project_name='test-project
    )

    # Wait until model is being uploaded.