Currently the only way to manually update the CRF state (such as start_transitions) is by accessing its internal memory directly (crf.start_transitions.data = my_new_parameters). Let's consider if there's a better way of achieving this - either through a setter function (see #18 ) or by allowing the parameters to be specified in the model constructor.
Currently the only way to manually update the CRF state (such as start_transitions) is by accessing its internal memory directly (crf.start_transitions.data = my_new_parameters). Let's consider if there's a better way of achieving this - either through a setter function (see #18 ) or by allowing the parameters to be specified in the model constructor.