Grigorios Kalliatakis

Update on my question above: training works ok when the sequential model is used instead, but still no with the functional model I shared above.

Hi @sazya have you found out if the [bayesian_categorical_crossentropy](https://github.com/kyle-dorman/bayesian-neural-network-blogpost/blob/5fc764b18e4826b90b1082e5142fdbbb23b0395d/bnn/loss_equations.py#L58) makes a double count of the softmax_output? Regarding the `from_logits` being set to True (taken from Keras doc) --> `from_logits: Boolean,...

The loss function is exactly the one described in this repo: ```python # Bayesian categorical cross entropy. # N data points, C classes, T monte carlo simulations # true -...

Have you solved the issue @Yapeng-Wang ?

Weights of the trained models for segmentation and EF prediction have already been released [here](https://github.com/echonet/dynamic/releases/tag/v1.0.0) For example, to load the weights into the EF model: ```python EJECTION_FRACTION_WEIGHTS_PATH = 'https://github.com/echonet/dynamic/releases/download/v1.0.0/r2plus1d_18_32_2_pretrained.pt' weights_destination_dir...

@chrilouk From the error message I guess there's a mismatch between the model architecture loaded from the torchvision module and the pretrained weights that you are trying to load into...