Closed gmgreg closed 3 years ago
I would imagine this error comes out when trying to train the network / do the forward pass, right?
Anyway, it is hard to debug without the actual sample that it failed on. However, looking at the traceback, make sure that there is no NaNs in your data. If that is not helpful, just try to find a single X
(of shape (n_samples, n_channels, lookback, n_assets)
) such that network(X)
gives you the above error and post it here!
Not sure what the issue is/was. I was able to avoid the issue by re-writing my code. I also eliminated raw_to_Xy
in my data loading. At this point I've been able to increase my dataset size and run through BachelierNet. It appears to be functional but I'm not fully understanding some of the nuances of DeepDow yet. More questions to come.
Thank you again for your patience.
Perfect!
I'm able to get the out-of-the box examples to execute successfully (getting_started and iid) when using the generated data but when using differenty toy datasets I get an AssertionError in the cvxcpy module.
Steps to reproduce:
Same error occurs even if reducing channels to 1, increasing number of samples, keeping lookback, gap, horizion small (5,0,1).