tallamjr / astronet

Efficient Deep Learning for Real-time Classification of Astronomical Transients and Multivariate Time-series
Apache License 2.0
14 stars 3 forks source link

Reimplement custom log loss as a tf.keras object #41

Closed tallamjr closed 3 years ago

tallamjr commented 3 years ago

When attempting to run evaluation and predictions on a loaded model that contains the custom loss, several errors occur.

It seems that from https://github.com/keras-team/keras/issues/5916 https://github.com/keras-team/keras/issues/5916#issuecomment-457624404 one needs to point to the custom_object. Currently the implementation is not in this form.