load_custom_hf_dataset not handling the text_feature argument properly

If you use a hf_dataset configuration such as (for e.x.)

hf_dataset:
  name: "Open-Orca/OpenOrca"
  train_split: "train[:90%]"
  valid_split: "train[-10%:]"
  text_feature: "response"

It is supposed to work the same as the (local) text data format, but it comes up against:

        if prompt_feature and completion_feature:
            return CompletionsDataset(ds, tokenizer, prompt_feature, completion_feature)
        elif text_feature:
            return Dataset(train_ds, text_key=text_feature)
        else:
            raise ValueError(
                "Specify either a prompt and completion feature or a text "
                "feature for the Hugging Face dataset."
            )

which errors out because train_ds is not defined

Search for load_custom_hf_dataset and inspect the branch shown in the issue, especially the text_feature path that references train_ds. Compare it with the local text data format and verify the Hugging Face configuration using Open-Orca/OpenOrca; done means text_feature loads the selected dataset without the undefined-variable error.