Downloading Waterbirds isn't Working Anymore

[divyansha]

Earlier, I was able to download Waterbirds, but now I'm not able to do so. Here's the traceback:

Traceback (most recent call last): File "/Users/divyanshagarwal/Documents/UCLA/Capstone/evaluate_waterbirds.py", line 46, in <module> dataset = get_dataset(dataset="waterbirds", download=args.download_waterbirds) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Users/divyanshagarwal/anaconda3/lib/python3.11/site-packages/wilds/get_dataset.py", line 75, in get_dataset return WaterbirdsDataset(version=version, **dataset_kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Users/divyanshagarwal/anaconda3/lib/python3.11/site-packages/wilds/datasets/waterbirds_dataset.py", line 72, in __init__ metadata_df = pd.read_csv( ^^^^^^^^^^^^ File "/Users/divyanshagarwal/anaconda3/lib/python3.11/site-packages/pandas/io/parsers/readers.py", line 912, in read_csv return _read(filepath_or_buffer, kwds) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Users/divyanshagarwal/anaconda3/lib/python3.11/site-packages/pandas/io/parsers/readers.py", line 577, in _read parser = TextFileReader(filepath_or_buffer, **kwds) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Users/divyanshagarwal/anaconda3/lib/python3.11/site-packages/pandas/io/parsers/readers.py", line 1407, in __init__ self._engine = self._make_engine(f, self.engine) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Users/divyanshagarwal/anaconda3/lib/python3.11/site-packages/pandas/io/parsers/readers.py", line 1661, in _make_engine self.handles = get_handle( ^^^^^^^^^^^ File "/Users/divyanshagarwal/anaconda3/lib/python3.11/site-packages/pandas/io/common.py", line 859, in get_handle handle = open( ^^^^^ FileNotFoundError: [Errno 2] No such file or directory: 'data/waterbirds_v1.0/metadata.csv'

[viictorjimenezzz]

I ran into the same issue, and it seems the URL is no longer functional. I recommend downloading the data manually and placing the extracted contents in the data/waterbirds_v1.0 directory. After that, the dataset can be initialized like this:

from wilds.datasets.waterbirds_dataset import WaterbirdsDataset

WaterbirdsDataset._versions_dict["1.0"]["download_url"] = None
dataset = WaterbirdsDataset(*args, **kwargs)

[HELL-TO-HEAVEN]

I ran into the same issue, and it seems the URL is no longer functional. I recommend downloading the data manually and placing the extracted contents in the data/waterbirds_v1.0 directory. After that, the dataset can be initialized like this:

from wilds.datasets.waterbirds_dataset import WaterbirdsDataset

WaterbirdsDataset._versions_dict["1.0"]["download_url"] = None
dataset = WaterbirdsDataset(*args, **kwargs)

@viictorjimenezzz U are right!!! Thanks!!!