chrismessenger

Why is the analysis run for a second time on line 388?

Best to remove all VICI references

Lines 618 to 621 seem to do a similar thing twice overwriting the first thing. It puts the r2 means and scales back into a single object but one version...

I thought that I'd changed this to be less ugly so that you can just call TruncatedNormal once over the appropriate parameters and then evaluate the log_prob and sum over...

On line 571 you probably want to have an else statement that sets the reconstruction_loss_gauss = 0.0 if gauss_len is not >0.

Why is the reconstruction loss summed on line 581 over the 1st and 2nd elements of the 2nd dimension? It looks like this a hard coded assumption that there are...

No need to define batch_norm since it is never used.

When normalising the data to be within the bounds and then scaling to be on between 0 and 1, it would be safe to check if the data is actually...

I think that the train_file_idx variable is generated assuming that there are 1000 time series stored per training data file. This may be correct (now) but should be fixed to...

The name given to the hidden layer batchnorm is not correct. It is being labelled with the same name as used in the conversation layer batchnorm. *If* this is a...