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

Marco Zucchini

08/30/2023, 8:34 PM
Dear Nixtla-Team, just starting with your fantastic library with a simple LSTM model: I have assured that the column 'ds' has a datetime.date format (pd.to_datetime). While the simple training (nf.fit) works, I get an error during training if I do a cross validation:
nf.cross_validation(df=Y_train_df, n_windows=10,step_size=1,val_size=horizon,sort_df=True,verbose=True)
`ValueError: You are trying to merge on datetime64[s] and object columns. If you wish to proceed you should use pd.concat`Actually I was not trying to merge anything at that point: is the method expecting another data format for 'ds' different from the simple fit? Thank you!
c

Cristian (Nixtla)

08/30/2023, 8:40 PM
Hi @Marco Zucchini! Can you show an example of the date? Can you try with
df['ds']=<http://pd.to|pd.to>_datetime(df['ds'])
?
m

Marco Zucchini

08/31/2023, 7:31 PM
Hi Cristian, thank you for quick reply! I was using:
df['ds']=<http://pd.to|pd.to>_datetime(df['Date'], utc=True).dt.date
, now I changed as you said but keeping
utc=True
and got now the date in the format e.g. 2021-08-31 040000+00:00. That error is not there anymore but somehow I have a new trouble (!)
CUDA unknown error - this may be due to an incorrectly set up environment
For the sake of overcome it quickly I have tried to switch to cpu but it does not work: how can I force cpu instead of gpu? not like that?
torch.cuda.is_available = lambda : False
Thanks!