Identical keywords in build_kwargs and config_kwargs lead to TypeError in load_dataset_builder()

Describe the bug

In load_dataset_builder(), build_kwargs and config_kwargs can contain the same keywords leading to a TypeError("type object got multiple values for keyword argument “xyz”).

I ran into this problem with the keyword: base_path. It might happen with other kwargs as well. I think a quickfix would be

builder_cls = import_main_class(dataset_module.module_path)
builder_kwargs = dataset_module.builder_kwargs
data_files = builder_kwargs.pop("data_files", data_files)
config_name = builder_kwargs.pop("config_name", name)
hash = builder_kwargs.pop("hash")
base_path = builder_kwargs.pop("base_path")

and then pass base_path into builder_cls.

Steps to reproduce the bug

from datasets import load_dataset
load_dataset("rotten_tomatoes", base_path="./sample_data")

Expected results

The docs state: **config_kwargs — Keyword arguments to be passed to the BuilderConfig and used in the DatasetBuilder.

So I would expect to be able to pass the base_path into load_dataset().

Actual results

TypeError("type object got multiple values for keyword argument “base_path”).

Environment info

• datasets version: 2.4.0
• Platform: macOS-12.5-arm64-arm-64bit
• Python version: 3.8.9
• PyArrow version: 9.0.0

[lhoestq]

Ok I see - maybe we should check the values of builder_kwargs raise an error if any key in config_kwargs tries to overwrite it ? The builder kwargs are determined from the builder’s type and location (in some cases it forces the base_path, data_files and config name for example)

[thepurpleowl]

I am getting similar error - TypeError: type object got multiple values for keyword argument 'name' while following this tutorial. I am getting this error with the dataset-cli test command.

datasets version: 2.4.0