-
Notifications
You must be signed in to change notification settings - Fork 660
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
TimeSeries API Bugs (frequency, context length, FEAT_DYNAMIC_REAL) #3271
Comments
also, in gluonTS cardinality can be "auto" or "ignore", if we dont use categorical features, but TimeSeries API docs doesnt indicate how can we do the same thing here |
also, in gluonTS we can make context length and prediction length different, but when i try to do it in TimeSeries API i get exception in any case, regardless of the length of the time series:
that is TrainTimeSeries.class from examples with some changes:
|
also got a strange exception when
in gluonTS feat_dynamic_real can be used without static_real |
@frankfliu also DeepARNetwork class has a few strange lines of code like TRAIN_INPUT_FIELDS or this |
Address part of issue deepjavalibrary#3271
Fixing part of deepjavalibrary#3271
If we specify the frequency as required by the Lag class or as the GluonTS documentation says,
for example "1H", "H", "15min", we will get an exception:
or
if we specify the frequency in the format required by the Period and Duration classes (for example "1h", "5h") we will receive an exception from DJL:
if we specify the frequency like "15M" = it will be 15 months, so with weeks and months everything is ok
The text was updated successfully, but these errors were encountered: