Closed azagniotov closed 3 years ago
This doesn't affect the training. You can pop
the feature stars
from the input features when constructing the labels.
@xuanhuiwang thank you. Right, but the label is already being pop
ed in the following code as per the aforementioned example:
features = tf.compat.v1.data.make_one_shot_iterator(dataset).get_next()
label = tf.squeeze(features.pop(_LABEL_FEATURE), axis=2)
label = tf.cast(label, tf.float32)
but, thank you again for confirming about the training,
I am am using the following example to train a model using a toy data (I can provide ELWC structure if needed): https://github.com/tensorflow/ranking/blob/master/tensorflow_ranking/examples/keras/keras_m2e_tfrecord.py
on the following TF stack:
During training I see the following warning about a missing key
stars
and and I do not know if this is a warning I should worry about. Thestars
is the name of my label feature in ELWCCould you please advise?