PyTorch Lightning

mlop provides super performant integration with PyTorch Lightning, including dedicated support for checkpointing, logging artifacts, model details, and more.

Migrating from Weights & Biases

See the Migrating from Weights & Biases guide for a quickstart.

Logging Lightning Models

To use mlop as a PyTorch or Fabric Logger for Lightning, simply import the logger class dedicated for Lightning and use it in the Trainer or Fabric initialization.

from mlop.compat.lightning import MLOPLogger
 
logger = MLOPLogger(
    op=None,
    project=None,
    name=None,
    **kwargs,
)
trainer = Trainer(logger=logger)
fabric = L.Fabric(loggers=[logger])

Parameter	Type	Description
`op`	`mlop.op.Op`	The initialized mlop run to log to.
`project`	`Optional[str]`	The name of the project (if uninitialized).
`name`	`Optional[str]`	The name of the run (if uninitialized).
`**kwargs`	`Any`	Additional keyword arguments accepted by the `mlop.init` constructor.

Source

See the Python code for more details.

Using `MLOPLogger`

The MLOPLogger inherits from the standard lightning.pytorch.loggers.Logger class, and strictly conforms to its standard API interface.

Method	Description	Example
`log_metrics`	Log metrics to mlop.	`logger.log_metrics(data: Dict[str, float])`
`log_hyperparams`	Log hyperparameters to mlop.	`logger.log_hyperparams(params: Dict[str, Any])`
`log_file`	Log a file to mlop.	`logger.log_file(key: str, files: List[mlop.File], **kwargs)`
`log_image`	Log an image to mlop.	`logger.log_image(key: str, images: List[mlop.Image], **kwargs)`
`log_audio`	Log an audio file to mlop.	`logger.log_audio(key: str, audios: List[mlop.Audio], **kwargs)`
`log_video`	Log a video to mlop.	`logger.log_video(key: str, videos: List[mlop.Video], **kwargs)`
`log_checkpoint`	Log a checkpoint to mlop.	`logger.log_checkpoint(key: str, checkpoint: "ModelCheckpoint", **kwargs)`
`log_graph`	Log a model graph to mlop.	`logger.log_graph(model: "torch.nn.Module", **kwargs)`
`watch`	Watch a model.	`logger.watch(model: "torch.nn.Module", **kwargs)`

PyTorch Lightning

Logging Lightning Models

Using MLOPLogger

On this page

Using `MLOPLogger`