__init__() got an unexpected keyword argument 'input_size'

### Describe the bug

I try to define a custom configuration with a input_size attribute following the instructions by "Specifying several dataset configurations" in https://huggingface.co/docs/datasets/v1.2.1/add_dataset.html

But when I load the dataset, I got an error "__init__() got an unexpected keyword argument 'input_size'"

### Steps to reproduce the bug

Following is the code to define the dataset:
class CsvConfig(datasets.BuilderConfig):
"""BuilderConfig for CSV."""
input_size: int = 2048

class MRF(datasets.ArrowBasedBuilder):
"""Archival MRF data"""
BUILDER_CONFIG_CLASS = CsvConfig
VERSION = datasets.Version("1.0.0")
BUILDER_CONFIGS = [
CsvConfig(name="default", version=VERSION, description="MRF data", input_size=2048),
]

...

def _generate_examples(self):
input_size = self.config.input_size
if input_size > 1000:
numin = 10000
else:
numin = 15000

Below is the code to load the dataset:
reader = load_dataset("default", input_size=1024)

### Expected behavior

I hope to pass the "input_size" parameter to MRF datasets, and change "input_size" to any value when loading the datasets.

### Environment info

- `datasets` version: 2.5.1
- Platform: Linux-4.18.0-305.3.1.el8.x86_64-x86_64-with-glibc2.31
- Python version: 3.9.12
- PyArrow version: 9.0.0
- Pandas version: 1.5.0