Experiment Management

LitLogger is the experiment tracker built into Lightning AI. It supports standalone scripts, the new dict-style Experiment API, Lightning/Fabric integration, files, media, models, and retrieval workflows.

Comparing experiment metrics with charts

These docs are organized as a narrative:

tutorials to get from zero to a working flow
guides for specific workflows and API patterns
runnable examples from the repository
an API reference for the public surface

Start Here

Use this path if you are new to LitLogger:

Supported Workflows

LitLogger currently supports these main workflows:

standalone logging with the dict-style Experiment API
legacy module-level logging helpers for existing scripts
Lightning and Fabric integration through upstream lightning.pytorch.loggers.LitLogger
file, image, text, and model logging through File, Image, Text, and Model
experiment resume and later retrieval by name through litlogger.init.init()
inference logging from a LitServe endpoint

Quick Example

import litlogger
from litlogger import File, Text

experiment = litlogger.init(name="quickstart")

experiment["model"] = "resnet50"
experiment["notes"] = Text("first run")

for step in range(10):
    experiment["train/loss"].append(1.0 / (step + 1), step=step)

experiment["config"] = File("config.yaml")
experiment.finalize()

Documentation Map

Tutorials provide end-to-end, task-oriented walkthroughs.

Guides explain workflows, migration paths, and logging patterns in more depth.

Examples include every runnable example currently shipped in the repository.

The API reference documents the public classes, functions, and enums.

Start

Installation

Tutorials

Guides

Examples

See Examples for the full list of runnable repository examples.

API Reference

API Reference

Was this page helpful?