[BUG] Valid split autoset to None for LLM training

[rhingo]

Prerequisites

• [X] I have read the documentation.
• [X] I have checked other issues for similar problems.

Backend

Local

Interface Used

CLI

CLI Command

autotrain --config <path_to_config.yml>

UI Screenshots & Parameters

No response

Error Logs

Not necessarily an error, but I noticed that for the LLM training task, in the llm_munge_data function in autotrain.cli.utils.py the params.valid_split by default is set to None. This overrides any user specified validation split in the yaml config file. For the other "munge_data" functions it looks like the default for params.valid_split = "validation". Can this please be fixed to allow the user to specify a validation split?

Additional Information

No response