ForecastDFDataset constructor mismatch in Hugging Face Blog, Tutorial and Codebase.

[arc-arnob]

In the tutorial and hugging face blog for PatchTST, we have:

ForecastDFDataset(
    time_series_preprocessor.preprocess(train_data),
    id_columns=id_columns,
    input_columns=forecast_columns,
    output_columns=forecast_columns,
    context_length=context_length,
    prediction_length=forecast_horizon,
)

But there are no input_columns and output_columns in the class construction definition.

Am I missing something here? Does this also cause downstream issues?

[arc-arnob]

I suppose they got replaced by observable_columns and target_columns but would like to see if the author agrees as well. Cheers 😄

[wgifford]

@arc-arnob Thanks for your interest in granite-tsfm. I will try to update that blog post to reflect changes in the API. For the example in the blog, I believe we need to replace input/output with just a single parameter: target_columns.