Use load_dataset to load imagenet-1K But find a empty dataset

### Describe the bug

```python
def get_dataset(data_path, train_folder="train", val_folder="val"):
traindir = os.path.join(data_path, train_folder)
valdir = os.path.join(data_path, val_folder)

def transform_val_examples(examples):
transform = Compose([
Resize(256),
CenterCrop(224),
ToTensor(),
])
examples["image"] = [transform(image.convert("RGB")) for image in examples["image"]]
return examples

def transform_train_examples(examples):
transform = Compose([
RandomResizedCrop(224),
RandomHorizontalFlip(),
ToTensor(),
])
examples["image"] = [transform(image.convert("RGB")) for image in examples["image"]]
return examples

# @fengsicheng: This way is very slow for big dataset like ImageNet-1K (but can pass the network problem using local dataset)
# train_set = load_dataset("imagefolder", data_dir=traindir, num_proc=4)
# test_set = load_dataset("imagefolder", data_dir=valdir, num_proc=4)

train_set = load_dataset("imagenet-1K", split="train", trust_remote_code=True)
test_set = load_dataset("imagenet-1K", split="test", trust_remote_code=True)

print(train_set["label"])

train_set.set_transform(transform_train_examples)
test_set.set_transform(transform_val_examples)

return train_set, test_set
```
above the code, but output of the print is a list of None:

### Steps to reproduce the bug

1. just ran the code
2. see the print

### Expected behavior

I do not know how to fix this, can anyone provide help or something? It is hurry for me

### Environment info

- `datasets` version: 2.21.0
- Platform: Linux-5.4.0-190-generic-x86_64-with-glibc2.31
- Python version: 3.10.14
- `huggingface_hub` version: 0.24.6
- PyArrow version: 17.0.0
- Pandas version: 2.2.2
- `fsspec` version: 2024.6.1