tensorflow - {TypeError}Using a `tf.Tensor` as a Python `bool` is not allowed -

i create tfrecord of sequences, , trying parse using:

context, sequence = tf.parse_single_sequence_example(         serialized_example,          context_keys_to_features,  # {'label': fixedlenfeature(shape=[], dtype=tf.int64, default_value=none)}         sequence_keys_to_features  # {'token': fixedlensequencefeature(shape=[], dtype=tf.int64, allow_missing=false, default_value=none)}         ) 

if try evaluate value of context , sequence, following:

tf.contrib.learn.run_n(sequence)  >> {'token': array([   17,    10,     6,   878,    25,  ... ])} 

same context:

tf.contrib.learn.run_n(context)  >> {'label': 1} 

but when try, evaluate tensor inside the dict, raises error:

# next line raise error tf.contrib.learn.run_n(sequence['token']) # line tf.contrib.learn.run_n(context['label'])  >> {typeerror}using `tf.tensor` python `bool` not allowed. use `if t not none:` instead of `if t:` test if tensor defined, , use tensorflow ops such tf.cond execute subgraphs conditioned on value of tensor. 

any ideas why case, checked whole pipeline, , don't seem use if instead of tf.cond or other operations raise condition.