Hi team. I am experimenting with neuralforecast for my use case and I am following the End to End Walkthrough from the documentation. When I am on step 4 and I am trying to 'Train multiple models for many time series', I am getting an error during

.fit

method. The error is related to Ray Tune and is as below. Is there any step I am missing?

FileNotFoundError: [WinError 3] The system cannot find the path specified: 'C:\\Users\\makar\\AppData\\Local\\Temp\\ray\\session_2024-03-26_15-59-12_264246_43048\\artifacts\\2024-03-27_11-47-38\\_train_tune_2024-03-27_11-47-38\\driver_artifacts\\_train_tune_d2386_00000_0_batch_size=1,input_size=144,learning_rate=0.0021,max_steps=1000,n_freq_downsample=1_1_1_1_1,n_pool_kerne_2024-03-27_11-47-38'

Please note that I am running this on my local. Additionally, the documentation says that if no

config

is provided to the Auto models, they will search a pre-defined search space for the hyper-parameters. However, I am getting the above error when I try to pass no config values as well. Thanks in advance!

This is a problem with Ray and Windows, it'll be fixed in the next release which we'll do in the next couple of days. You can set

backend='optuna'

in the meantime.

hey @José Morales, as you said I am following the tutorial for hyper-parameter optimisation using Optuna and the tutorial mentions the use of

AutoNHITS.get_default_config(h = 12, backend="optuna")

This returns the below error

AttributeError: type object 'AutoNHITS' has no attribute 'get_default_config'

That's part of the new release (1.7.0), you should be able to use it by upgrading or you can just run AutoNHITS(h=12, backend='optuna')