Introduction

Within mlop, experiments are the main unit of work. They are created by running a script with the mlop client.

Creating an experiment

To create an experiment, you simply start the mlop client with the .init() function with the name of the experiment name and the project project.

import mlop
 
mlop.init(name="gpt4o-vision", project="VLLM")

You can pass in configurations, e.g hyperparameters the config parameter

hyperparameters = {
    "lr": 0.001,
    "epochs": 1000,
}
 
mlop.init(
    dir=".",
    name="gpt4o-vision",
    project="VLLM",
    config=hyperparameters,
)

and this will be recorded in the experiment.

mlop.Settings

One can also pass in the settings parameter to change the configuration of the logger. You most likely don't want to do this unless you are debugging the logger or self-hosting the application.

Parameter	Type	Description
`dir`	`str`	The directory to save the experiment.
`name`	`str`	The name of the experiment.
`project`	`str`	The project to save the experiment.
`config`	`dict`	The hyperparameters/configuration for the experiment.
`settings`	`dict`	The logger settings for the experiment.

This will create an experiment with the name gpt4o-vision and the project VLLM. To see more about projects, check out the projects page.

Introduction

Creating an experiment

On this page