Isaac Godfried
Isaac Godfried
It looks like DilateLoss is correlated with MSE when `forecast_length == 1`, however the tensor might be in the improper dimension when it is passed to dilate loss. In any...
I think they likely added additional checks on data dimensions which is why this is failing.
The fix for this pretty easy. Essentially just replace with torch.arrange() `This function is deprecated and will be removed in a future release because its behavior is inconsistent with Python’s...
It think this is more a limitation of DilateLoss that it cannot work with multiple time series at once. Instead we could add code to DilateLoss to run it each...
So currently these features are not supported. We hope to have support for these features in the next release 4/15.
Relevant notebook: https://colab.research.google.com/drive/1mbn777MigWa-Q6WbUnhZ0OxVTHcO7lee#scrollTo=EMAcDy8ufyAX Scenario: happens when sweep_config has ``` meta_data: { "values":[True, False] } ``` I'm still wondering why though. My guess would be a part of the initialized mode...
@joelrorseth Could you describe your problems in more detail? As for your question above the `make_final_config` aims to solve the problem of creating a flow-forecast valid configuration file from weights...
@chrisoyer
Yes I'd be interested in this as well.