plot_forecast function not working with alternate energy price forecast

[oavaldezi]

I'm using the Energy Price Forecasting example in the chronos-2-quickstart notebook to forecast prices for a different energy market. The energy_context_df, energy_test_df and energy_future_df dataframes seem to load correctly, and predictions are generated:

display(energy_context_df.tail()):

id	timestamp	Gen1	Gen2	Gen3	Cost1	Cost2	target
XX	2025-11-22 19:00:00	687.60	358.52	94.14	9.746	70.012	75.746
XX	2025-11-22 20:00:00	688.26	355.78	68.33	9.746	70.012	75.746
XX	2025-11-22 21:00:00	652.02	354.99	46.97	9.746	70.012	75.746
XX	2025-11-22 22:00:00	474.06	356.44	46.85	9.746	70.012	99.039
XX	2025-11-22 23:00:00	439.07	342.21	35.74	9.746	70.012	99.039

display(energy_future_df.head()):

id	timestamp	Gen1	Gen2	Gen3	Cost1	Cost2
XX	2025-11-23 00:00:00	385.46	334.79	14.58	9.746	70.012
XX	2025-11-23 01:00:00	382.37	323.16	0.00	9.746	70.012
XX	2025-11-23 02:00:00	384.11	300.70	0.00	9.746	70.012
XX	2025-11-23 03:00:00	345.61	326.26	0.00	9.746	70.012
XX	2025-11-23 04:00:00	323.45	335.36	0.00	9.746	70.012

display(energy_pred_df.head()):

id	timestamp	target_name	predictions	0.1	0.5	0.9
XX	2025-11-23 00:00:00	target	87.661774	80.629868	87.661774	92.886375
XX	2025-11-23 01:00:00	target	76.276794	67.580238	76.276794	82.534935
XX	2025-11-23 02:00:00	target	75.555023	66.740967	75.555023	81.084442
XX	2025-11-23 03:00:00	target	75.894638	68.155273	75.894638	80.635788
XX	2025-11-23 04:00:00	target	75.680847	67.562347	75.680847	81.126083

However, forecasts aren't plotted as in the example:

The historical target, the future target (ground truth), the forecast and the prediction interval seem to be bunched together at the left, and the only timestamp shown is the first hourly interval in the prediction period, instead of the dates of of the historical period, as in the example. What could be happening?

[abdulfatir]

It looks like you're using a custom dataset here, not the energy price from the quick start notebook. Please check if you are passing all arguments to the plotting function correctly:

def plot_forecast(
    context_df: pd.DataFrame,
    pred_df: pd.DataFrame,
    test_df: pd.DataFrame,
    target_column: str,
    timeseries_id: str,
    id_column: str = "id",
    timestamp_column: str = "timestamp",
    history_length: int = 256,
    title_suffix: str = "",
):

[kabaros]

I had a similar issue with plot_forecast with an external dataset.

For some reason, when plotting but not when predicting, I had to convert the timestamp fields to datetime again, so something like

energy_context_df['timestamp'] = pd.to_datetime(energy_context_df['timestamp'])
energy_future_df['timestamp'] = pd.to_datetime(energy_future_df['timestamp'])

then pass the data frames to plot_forecast

(it wasn't necessary for the predictions dataset in my case, but looking at your screenshot, it might be necessary to do the same transformation on that dataframe)

Before	After

[abdulfatir]

Makes sense! Thanks for sharing.

[oavaldezi]

I'm importing my external dataset from a CSV file, and the timestamp field was being imported with dtype object. Upon @kabaros' suggestion, these two lines convert the timestamps in energy_context_df and energy_test_df to datetime64[ns] (energy_future_df doesn't need fixing because its timestamp is generated with dtype datetime64[ns]):

energy_context_df['timestamp'] = pd.to_datetime(energy_context_df['timestamp'])
energy_test_df['timestamp'] = pd.to_datetime(energy_test_df['timestamp'])

The figures then display correctly.