https://github.com/nixtla logo
#neural-forecast
Title
# neural-forecast
a

Asterios Tsiourvas

04/17/2023, 6:28 PM
Hello. I receive the following error when running the NBEATS model with horizon = 1. RuntimeError: einsum(): subscript p has size 2 for operand 1 which does not broadcast with previously seen size 0 Is there a fix on that? Thanks!!
c

Cristian (Nixtla)

04/18/2023, 12:32 AM
Hi @Asterios Tsiourvas! Are you using the defaults for the other hyperparameters? What input_size are you using?
a

Asterios Tsiourvas

04/18/2023, 10:20 PM
Thank you for the response! I am using nbeats_config = { “input_size”: tune.choice([2*seasonality]), “learning_rate”: tune.loguniform(1e-4, 1e-1), “scaler_type”: tune.choice([None, “robust”, “standard”]), “max_steps”: tune.choice([500, 1000]), “batch_size”: tune.choice([32, 64, 128, 256]), “windows_batch_size”: tune.choice([128, 256, 512, 1024]), “random_seed”: tune.randint(1, 20), “val_check_steps”: tune.choice([1]), “early_stop_patience_steps”: tune.choice([10]), “max_epochs”: tune.choice([100]) } models=[ AutoNBEATS(h=horizon, config = nbeats_config, loss = MAE(), refit_with_val=True, num_samples=10) ] and the Wiki2 dataset from the hierarchicalforecast package.
c

Cristian (Nixtla)

04/18/2023, 11:05 PM
I will try to replicate the error
a

Asterios Tsiourvas

04/19/2023, 10:21 PM
Thank you!
14 Views