Hi TFP team,
In the TFP tutorial, the model output is Normal distribution. I noted that the output can be replaced by an IndependentNormal layer. In my model, the y_true is binary class. Therefore, I used an IndependentBernoulli layer instead of IndependentNormal layer.
After building the model, I found that it has two output parameters. It doesn’t make sense to me since Bernoulli distribution has one parameter only. Do you know what went wrong?