TensorFlow DataSet API causes graph size to explode -
i have bug data set training.
i'm using data set api so:
self._dataset = tf.contrib.data.dataset.from_tensor_slices((self._images_list, self._labels_list)) self._dataset = self._dataset.map(self.load_image) self._dataset = self._dataset.batch(batch_size) self._dataset = self._dataset.shuffle(buffer_size=shuffle_buffer_size) self._dataset = self._dataset.repeat() self._iterator = self._dataset.make_one_shot_iterator()
if use training small amount of data well. if use data tensorflow crash error: valueerror: graphdef cannot larger 2gb.
it seems tensorflow tries load data instead of loading data needs... not sure...
any advice great!
update... found solution/workaround
according post: tensorflow dataset api doubles graph protobuff filesize
i replaced make_one_shot_iterator() make_initializable_iterator() , of course called iterator initializer after creating session:
init = tf.global_variables_initializer() sess.run(init) sess.run(train_data._iterator.initializer)
but i'm leaving question open me seems workaround , not solution...
Comments
Post a Comment