Training File Edge Case Handling

[Ataraxist]

Problem

Right now the system ONLY accepts .arrow files, but training data may be in parquets or JSONL, or some other data type. The tkyo train command should be able to accept a more diverse file types, and should be able to handle more object structures within those file types.

Solution

The Resolve Column IO python function needs to be expanded for edge cases.

Additional information

No response

👨‍👧‍👦 Contributing

• 🙋‍♂️ Yes, I'd love to make a PR to implement this feature!