---
title: Lightning Integration
description: available only as a deprecated compatibility alias.
canonical: 'https://lightning.ai/docs/litlogger/stable/guides/lightning'
---
# Lightning Integration {#lightning-integration}

Use `lightning:lightning.pytorch.loggers.LitLogger` for Trainer and
Fabric integration. [`LightningLogger`](/docs/litlogger/2026.08.28/api-reference/api.md#litlogger.logger.LightningLogger) remains
available only as a deprecated compatibility alias.

## Using with Trainer {#using-with-trainer}

Pass `lightning:lightning.pytorch.loggers.LitLogger` as the `logger`
argument to the Trainer. Every `self.log()` call inside your LightningModule
is forwarded automatically.

```python
from lightning import Trainer
from lightning.pytorch.loggers import LitLogger

logger = LitLogger(name="cifar10-resnet")
trainer = Trainer(max_epochs=10, logger=logger)
trainer.fit(model, datamodule)
```

After `trainer.fit()` starts, the logger prints a URL where you can view
live training curves.

## Logging Hyperparameters {#logging-hyperparameters}

Lightning automatically calls `log_hyperparams` when your LightningModule
defines `self.save_hyperparameters()`.

```python
class MyModel(L.LightningModule):
    def __init__(self, lr: float = 1e-3, hidden_dim: int = 128):
        super().__init__()
        self.save_hyperparameters()
```

You can also pass metadata directly:

```python
logger = LitLogger(
    name="cifar10-resnet",
    metadata={"optimizer": "AdamW", "scheduler": "cosine"},
)
```

## Automatic Checkpoint Logging {#automatic-checkpoint-logging}

Set `log_model=True` to automatically upload checkpoints to the litmodels
registry whenever Lightning saves a checkpoint.

```python
from lightning.pytorch.callbacks import ModelCheckpoint

checkpoint_cb = ModelCheckpoint(save_top_k=2, monitor="val_loss")
logger = LitLogger(name="my-model", log_model=True)

trainer = Trainer(
    max_epochs=20,
    logger=logger,
    callbacks=[checkpoint_cb],
)
trainer.fit(model, datamodule)
```

## Using with Fabric {#using-with-fabric}

`lightning:lightning.pytorch.loggers.LitLogger` also works as a Fabric
logger:

```python
import lightning as L
from lightning.pytorch.loggers import LitLogger

logger = LitLogger(name="fabric-run")
fabric = L.Fabric(loggers=[logger])
fabric.launch()

for step in range(1000):
    loss = train_step()
    fabric.log("train_loss", loss, step=step)
```

## Logging Files and Models {#logging-files-and-models}

The logger also exposes matching file and model logging helpers on
`lightning:lightning.pytorch.loggers.LitLogger`:

```python
trainer.logger.log_file("config.yaml")
trainer.logger.log_model_artifact("checkpoints/best.ckpt", version="best")
```

For media uploads in Lightning workflows, use the standalone
[`Experiment`](/docs/litlogger/2026.08.28/api-reference/api.md#litlogger.experiment.Experiment) API directly.

## Accessing the Experiment URL {#accessing-the-experiment-url}

Use the logger's `url` property to retrieve the run URL from the logger
instance.

```python
logger = LitLogger(name="my-run")
trainer = Trainer(logger=logger)
trainer.fit(model, datamodule)

print(logger.url)
```

---

**Navigation**

- Parent: [Guides](/docs/litlogger/2026.08.28/guides.md)