https://github.com/nixtla logo
r

Rajat Sen

07/05/2023, 7:39 PM
Hi everyone, Thanks for the great library. I was wondering if there is a way to distinguish between categorical features and numerical features within
Copy code
futr_exog: torch.Tensor,
hist_exog: torch.Tensor,
stat_exog: torch.Tensor,
so that we can embed the categorical features with learnable embeddings ? Also for categorical features we probably need to know the cardinality of the feature in the training set as well. What do you think?
k

Kin Gtz. Olivares

07/05/2023, 7:52 PM
Hey @Rajat Sen , we are not making a distinction between categorical and numeric features for the moment. We adapted TFT for that purpose here: https://github.com/Nixtla/neuralforecast/blob/main/neuralforecast/models/tft.py