multi-label classification error based on tf.slim - tensorflow

Question

When build tfrecord, I have encoded the labels as N-hot encoding.
Then I comment the following code:
labels = slim.one_hot_encoding(
labels, dataset.num_classes - FLAGS.labels_offset)
But it crash when train:
Traceback (most recent call last):
File "../train_image_classifier_multilabel.py", line 620, in <module>
tf.app.run()
File "/root/anaconda3/lib/python3.6/site-packages/tensorflow/python/platform/app.py", line 126, in run
_sys.exit(main(argv))
File "../train_image_classifier_multilabel.py", line 519, in main
clones = model_deploy.create_clones(deploy_config, clone_fn, [batch_queue])
File "/root/Projects/scene-tf/slim/deployment/model_deploy.py", line 193, in create_clones
outputs = model_fn(*args, **kwargs)
File "../train_image_classifier_multilabel.py", line 513, in clone_fn
label_smoothing=FLAGS.label_smoothing, weights=1.0)
File "/root/anaconda3/lib/python3.6/site-packages/tensorflow/python/ops/losses/losses_impl.py", line 676, in sigmoid_cross_entropy
logits.get_shape().assert_is_compatible_with(multi_class_labels.get_shape())
File "/root/anaconda3/lib/python3.6/site-packages/tensorflow/python/framework/tensor_shape.py", line 764, in assert_is_compatible_with
raise ValueError("Shapes %s and %s are incompatible" % (self, other))
ValueError: Shapes (32, 63) and (32,) are incompatible
The sigmoid_cross_entropy encounters problem:
tf.losses.sigmoid_cross_entropy(
logits=logits, multi_class_labels=labels,
label_smoothing=FLAGS.label_smoothing, weights=1.0)
Obviously, the shape of logits is (32, 63) and labels is (32,) .
It seems that the comment of one_hot_encoding is wrong.
But I don't know how to fix it.