$ ml --outputformat [tables|json]
This option control the structure of the output of any ml command the default is tables. In case you are using the MissingLink CLI as part of a 3rd-party script, we have added the JSON format for easier manipulation of the return values. The command has a shortcut of -o.
Model hash retrieval
MissingLink CLI can calculate your model's weights hash which can be used for more advanced integration like sending external metrics and charts.
ml models get-weights-hash command.
# For a Keras model ml models get-weights-hash --framework keras --model checkpoint.hdf5 # For a TensorFlow model ml models get-weights-hash --framework tf --model models/checkpoint.ckpt # For a Caffe model ml model get-weights-hash --framework caffe --model train.prototxt \ --weights checkpoints.caffemodel
For a full overview for this feature, see Source Tracking. You can establish source tracking by calling:
ml code track --source-repository=https://github.mock/myorg/myrepo [email protected]:myorg/myrepo_tracking.git
gitschemes are supported. - The mapping is available to any project started by an owner with access to the specified
ORG. - The tracking is enabled for every future experiment.
When setting up tracking using MissingLink CLI, both repositories MUST belong to the same organization. In the above example, an attempt to pass
[email protected]:notmyorg/myrepo_tracking.git as an argument will yield an error.