An example for finetuning FLAVA or any VLP multimodel using trainer (for example for classification)

### Feature request

There is no example of finetuning any VLP model using trainer. I would appreciate an example

### Motivation

The way to use trainers with any Vision and Language pretrained model is not clear.

### Your contribution

None.

Start by reviewing the existing Trainer examples and the FLAVA model entry points to determine how a vision-language classification fine-tuning example should be structured. Done means adding a clear, runnable example that demonstrates fine-tuning a VLP model with Trainer and explains the required data and configuration.

Written by the indexing model from the issue text.