AndiBerber comments

Results 4 comments of


                                            AndiBerber

Handle Constant Regions Consistently

@seanlaw : I guess it´s correct if the other one is random. But as it is the case now in `.mass()` it is set to `sqrt(m)` independent of the other...

AttributeError: 'SymbolicTensor' object has no attribute 'log_prob' when exporting train signature with `IndependentNormal` layer

Yes, ofc, stupid me. It works totally fine when i change that line. Thanks a lot.

AttributeError: 'SymbolicTensor' object has no attribute 'log_prob' when exporting train signature with `IndependentNormal` layer

```python loss = nll(predictions, targets) ``` should be ```python loss = nll(targets, predictions) ```

AttributeError: 'Tensor' object has no attribute 'log_prob'

The problem still occurs with tensorflow=2.15 and tensorflow-probability=0.23. None of the approaches in `https://github.com/tensorflow/probability/issues/742` works for me. Exporting for inference works, but the training signature throws with AttributeError: 'SymbolicTensor' object...