[Bug]: identify_dynamic_embeddings does not work for DataPair

[MattGPT-ai]

Describe the bug

The training util identify_dynamic_embeddings does not look properly at the embeddings for the first and second. It probably makes more sense to have a member function for each DataPoint class, instead of checking isinstance within an external function. There could still be an outer function that iterates over data points in a batch, but it should itself call identify_dynamic_embeddings_point for each data point it iterates over. I would say that identify_dynamic_embeddings should be called identify_dynamic_embeddings_batch, but that would change the API.

To Reproduce

first = flair.data.DataPoint()
second = flair.data.DataPoint()
dynamic_tensor1 = torch.tensor([1., 2., 3.], requires_grad=True)
dynamic_tensor2 = torch.tensor([1., 2., 3.], requires_grad=True)
first.set_embedding('dynamic', dynamic_tensor1)
second.set_embedding('dynamic', dynamic_tensor2)
dynamic_data_pair = flair.data.DataPair(first, second)
print(flair.training_utils.identify_dynamic_embeddings([dynamic_data_pair]))

Expected behavior

This should return ['dynamic']

Logs and Stack traces

Screenshots

No response

Additional Context

No response

Environment

Versions:

Flair

0.15.1

Pytorch

2.6.0+cu124

Transformers

4.48.2

GPU

False