tf.contrib.crf.crf_log_norm(inputs, sequence_lengths, transition_params)
See the guide: CRF (contrib)
Computes the normalization for a CRF.
inputs: A [batch_size, max_seq_len, num_tags] tensor of unary potentials to use as input to the CRF layer.
sequence_lengths: A [batch_size] vector of true sequence lengths.
transition_params: A [num_tags, num_tags] transition matrix. Returns:
log_norm: A [batch_size] vector of normalizers for a CRF.