I'm using the branch datapipeline_poc_1 with the following snippet to load a dataset and I got an error related to how the paths are specified.
data = ImageClassificationData.from_folders(
#train_folder=os.path.join(root, 'train'), # CRASHES in the assert below
train_folder=os.path.join(root),
train_transform = train_transform,
)
assert data._train_dataloader() is not None
The error is the following:
ValueError: num_samples should be a positive integer value, but got num_samples=0
My data tree in root is the following:
The loader is only able to fetch the data when the path passed to train_folder points to root instead of train.
In my opinion there's should be a mechanism that checks on a well formed or expected data tree.
edgarriba
commented
3 years ago
🐛 Bug
I'm using the branch datapipeline_poc_1 with the following snippet to load a dataset and I got an error related to how the paths are specified.
The error is the following:
ValueError: num_samples should be a positive integer value, but got num_samples=0My data tree in
rootis the following:The loader is only able to fetch the data when the path passed to
train_folderpoints torootinstead oftrain.In my opinion there's should be a mechanism that checks on a well formed or expected data tree.