In the guide, the pseudocode outputs just the Viterbi path, i.e., the best sequence of states. In the LXMLS code, the function "run_viterbi" from file sequences/sequence_classification_decoder.py, which is called from function "viterbi_decode" in file sequences/sequence_classifier.py, is expected to return the best sequence of states and the corresponding score:
In the guide, the pseudocode outputs just the Viterbi path, i.e., the best sequence of states. In the LXMLS code, the function "run_viterbi" from file sequences/sequence_classification_decoder.py, which is called from function "viterbi_decode" in file sequences/sequence_classifier.py, is expected to return the best sequence of states and the corresponding score:
best_states, total_score = self.decoder.run_viterbi(initial_scores, transition_scores, final_scores, emission_scores)
We should add, in the guide, that the output is the best path AND the corresponding score.