Example to train a lattice model with monotonic constraints
PoorvaRane opened this issue · comments
Hi! Could you please share an example to train a lattice with monotonic constraints?
I tried using projection ops but am facing an error while saving and restoring the model from the checkpoint (similar to Issue #35)
I assume you are using an estimator API. This is an example: https://github.com/tensorflow/lattice/blob/master/examples/coffee_test.py
Please share further details about the error you are getting, including your TF, TFL and python version.
On another note, I suggest you wait for the new library to be released, which uses Keras layers and handles constraints automatically. I expect that to happen in 3-4 weeks.
Thanks @mmilanifard ! I am using the estimator API and tfl.calibration_layer()
and tfl.lattice_layer()
for the calibrations and lattice. I am also using tf.estimator.train_and_evaluate
with tf.estimator.TrainSpec()
and tf.estimator.EvalSpec()
to train and evaluate my model. I suspect there may also be something to do with the tf.estimator.train_and_evaluate
function too.
My TF, TFL and python versions are:
python 2.7.10
tensorflow 1.14.0
tensorflow-lattice 0.9.9
Alright! I can wait for the new version to be released as well. Thanks!
Sorry for the long delays on this. RFC is out and we are hoping for a release by mid Jan.
PR: tensorflow/community#186
RFC doc