Add input data parameter to the encode_dataset.py

[ltsabadz]

Hi! Needed to follow docs couple of times and noticed that input data path was hardcoded as data/era5_240/full/. Would be helpful to have as an argument.

[robert-DL]

Hi, thx for the commit. In general, the encode_dataset.py is not really perfect as it also not allows for parallel generation or longer trajectories to be encoded .

[ltsabadz]

I also noticed that when you run the script on a subset of the data (e.g. 2007–2018), it still starts by saving files named as 1979 rather than the first year in the subset.

Another issue: if an older file like era5_240_pred_1979_0h.nc is already present in the output directory, and you try to run the script only for later years, the entire dataset gets skipped because of the following block:

current_year = 1979
xr_list = []
for i, batch in tqdm(enumerate(dl)):
    fname = Path(args.output_path).joinpath(f"era5_240_pred_{current_year}_0h.nc")
    if fname.exists():
        continue

But that's another topic..

[robert-DL]

The issue u mentioned here is actually already part of another issue.
Could you rebase to dev?

[ltsabadz]

The issue u mentioned here is actually already part of another issue. Could you rebase to dev?

Rebased and changed the base branch to dev