Enhancements to Dataset Processing and Inference Implementation

[gokceuludogan]

This pull request introduces enhancements to the dataset processor and new inference capabilities. Here is a summary of the key changes:

Key Changes

• An optional return_tensors has been added to the DatasetProcessor's tokenize_function to be used during inference.
• Documentation has been added to the Dataset Processor.
• A new predictor implementation for easy inference has been introduced.
• An inference notebook has been added.

Minor fixes:

• The installation command in notebooks has been fixed.
• Redundant imports from t5_classifier have been removed.

[gokceuludogan]

I believe using relative imports is more practical as it allows for flexibility for potential modifications and refactoring.

Plus, once the library is built, the structure of the imports, whether relative or absolute, becomes a non-issue for the end user.