diff --git a/src/crf_context.c b/src/crf_context.c
index 275454ee..372a15c9 100644
--- a/src/crf_context.c
+++ b/src/crf_context.c
@@ -564,6 +564,9 @@ double crf_context_viterbi(crf_context_t *self, uint32_t *labels) {
     const double *trans = NULL;
 
     const size_t T = self->num_items;
+    if (T == 0) {
+        return max_score;
+    }
     const size_t L = self->num_labels;
 
     // This function assumes state and trans scores to be in the logarithm domain.
diff --git a/src/crf_trainer_averaged_perceptron.c b/src/crf_trainer_averaged_perceptron.c
index 2f7cdeaa..ee3801a2 100644
--- a/src/crf_trainer_averaged_perceptron.c
+++ b/src/crf_trainer_averaged_perceptron.c
@@ -15,9 +15,8 @@ void crf_averaged_perceptron_trainer_destroy(crf_averaged_perceptron_trainer_t *
         kh_destroy(feature_class_weights, self->weights);
     }
 
+    khash_t(prev_tag_class_weights) *prev_tag_weights;
     if (self->prev_tag_weights != NULL) {
-        khash_t(prev_tag_class_weights) *prev_tag_weights;
-
         kh_foreach(self->prev_tag_weights, feature_id, prev_tag_weights, {
             if (prev_tag_weights != NULL) {
                 kh_destroy(prev_tag_class_weights, prev_tag_weights);
@@ -602,6 +601,10 @@ bool crf_averaged_perceptron_trainer_train_example(crf_averaged_perceptron_train
         return false;
     }
 
+    if (num_tokens == 0) {
+        return true;
+    }
+
     uint32_array_clear(self->sequence_features_indptr);
     uint32_array_push(self->sequence_features_indptr, 0);
     cstring_array_clear(self->sequence_features);