Describe the bug
When I try to export ner-large transformer embeddings to ONNX, I see an error TypeError: forward() missing 1 required positional argument: 'lengths'

To Reproduce

1. Install Flair==0.11.3 from git repo: pip install git+https://github.com/flairNLP/flair.git (TransformerOnnxEmbeddings class is not available when installed from pip)
2. Run code from the tutorial 13 with my own sentences:

from flair.models import SequenceTagger
from flair.embeddings import TransformerWordEmbeddings, TransformerDocumentEmbeddings

model = SequenceTagger.load("ner-large")
assert isinstance(model.embeddings, (TransformerWordEmbeddings, TransformerDocumentEmbeddings))


sentences = [Sentence(s) for s in ['short sentence', 'this is very long sentence']]
model = SequenceTagger.load("ner-large")
model.embeddings = model.embeddings.export_onnx("flert-embeddings.onnx", sentences, providers=["CUDAExecutionProvider", "CPUExecutionProvider"])

3. I get the error TypeError: forward() missing 1 required positional argument: 'lengths'

Expected behavior
This creates a file flert-embeddings.onnx which stores the exported Onnx Model.

Screenshots

Environment (please complete the following information):

• Deep Learning AMI GPU PyTorch 1.12.1 (Amazon Linux 2)
• NVIDIA driver version: 510.73.08
• CUDA version: 11.6
• Flair 0.11.3
• onnx 1.12.0
• onnxruntime 1.12.1
• Pytorch 1.12.1.post201