search

time-series-foundation-models / lag-llama

Lag-Llama: Towards Foundation Models for Probabilistic Time Series Forecasting
Apache License 2.0
1.09k stars 121 forks source link

How not to use "item_id"? #24

Closed sebasmos closed 3 months ago

sebasmos commented 3 months ago

I have a dataset with only 1 time series array, so item_id is always the same, which is equivalent to not to use it. On the Demo you mention: The item_id is required only when your dataset has multiple series. However its not working when not specifying this parameter, similarly when using PandasDataset(dfs_dict, target="target"). Do you have any suggestions on how to adapt the data on this case?

Thanks a lot

ashok-arjun commented 3 months ago

Hi! I haven't tested it with a single series; maybe it still needs an item_id - I'll check it.

For the time being, can you specify a random (constant) item_id in the item_id column, and see if it works? It shouldn't have any effect on the predictions; it is just for the PandasDataset.

shahrokhvahabi commented 3 months ago

I have a dataset with only 1 time series array, so item_id is always the same, which is equivalent to not to use it. On the Demo you mention: The item_id is required only when your dataset has multiple series. However its not working when not specifying this parameter, similarly when using PandasDataset(dfs_dict, target="target"). Do you have any suggestions on how to adapt the data on this case?

Thanks a lot

If an error appears you can set 'item_id' to 'A' just to justify the model you have just one time series array. I used it for just one array in 'df_wide' in the third section of the demo code and it works without 'item_id'.

kashif commented 3 months ago

i believe if you pull now you should not see this issue

ashok-arjun commented 3 months ago

Hi @sebasmos, can you confirm if the issue still exists please?

sebasmos commented 3 months ago

Yes thanks, the issue was also my own data, which I hadn't noticed it had some missing values. Part 3 of the demo works for me, thanks a lot!