Questions Regarding Operational Forecasts/Inference Based on GenCast 0p25deg Operational _2022.npz

Dear expert,

I have tried operational forecasts/inference based on GenCast 0p25deg Operational _2022.npz. The following are the questions that are troubling me — I would greatly appreciate your help in resolving them:

1. **Input Data Variable**: In the task_config, input_variables does not include **total_precipitation_12hr**. However, it seems that precipitation must be provided for inference. If using GFS data for operational inference, what kind of precipitation data should be used as input?

2. **Input Data Time Dimension**:
- In the example of gencast_demo_cloud_vm.ipynb, the input includes data at times t0, t1, and t2, while the output has a time dimension of 1. Is this corresponding to t2?
- For example, if I want to perform a forecast on June 13 at 12:00 for June 14–19, should I input data from June 13 at 00:00 and 12:00, along with data from June 14–19?
- How can I obtain the future data from June 14–19?
- And how should I modify the script to run such an inference?

Best regards

Start with gencast_demo_cloud_vm.ipynb and the task_config referenced in the issue. Trace the input_variables and time dimension used for inference, then determine the required precipitation and future-data inputs for the operational forecast scenario. Done means documenting clear answers to each question and identifying any necessary script changes.

Written by the indexing model from the issue text.