Where can I find the custom loss functions

[ezeid]

I am trying to load the lvh model: ecg_rest_raw_age_sex_bmi_lvm_asymmetric_loss.h5 with tensorflow
model = load_model(model_path), but it complains about a bunch of custom loss functions.

'symmetric_logcosh_loss_stable': symmetric_logcosh_loss_stable,
'loss_fxn': loss_fxn,
'asymmetric_outlier_mse':asymmetric_outlier_mse,
'lvh_from_indexed_lvm_weighted_loss': lvh_from_indexed_lvm_weighted_loss,
'no_lvh_precision': no_lvh_precision,
'left_ventricular_hypertrophy_precision': left_ventricular_hypertrophy_precision,
'no_lvh_recall': no_lvh_recall
...

I could not find these functions defined in the repo. Where can I find it?

[rxrathod]

Hi @ezeid , Use with compile false tag, you can load the model in this way. This should not give you any issues with custom functions

model = tf.keras.models.load_model("model.h5", compile=False)

cc: @lucidtronix