Skip to main content

MMCV v1.x

info

ClearMLLoggerHook is supported by mmcv =>1.5.1 and <=1.7.0.

tip

If you are not already using ClearML, see Getting Started for setup instructions.

MMCV is a computer vision framework developed by OpenMMLab. You can integrate ClearML into your code using the mmcv package's ClearMLLoggerHook class. This class is used to create a ClearML Task and to automatically log metrics.

For example, the following code sets up the configuration for logging metrics periodically to ClearML, and then registers the ClearML hook to a runner, which manages training in mmcv:

log_config = dict(
interval=100,
hooks=[
dict(
type='ClearMLLoggerHook',
init_kwargs=dict(
project_name='examples',
task_name='OpenMMLab cifar10',
output_uri=True
)
),
]
)

# register hooks to runner and those hooks will be invoked automatically
runner.register_training_hooks(
lr_config=lr_config,
optimizer_config=optimizer_config,
checkpoint_config=checkpoint_config,
log_config=log_config # ClearMLLogger hook
)

The init_kwargs dictionary can include any parameter from Task.init().

This creates a ClearML Task OpenMMLab cifar10 in the examples project. You can view the captured metrics in the experiment's Scalars tab in the WebApp.

OpenMMLab scalars

See MMCV code example here.