Add support for jinaai/jina-embeddings-v2-base-de

[deichrenner]

Issue

This PR resolves #266.

A new version of jina embeddings was added:

    {
        "model": "jinaai/jina-embeddings-v2-base-de",
        "dim": 768,
        "description": "German embedding model supporting 8192 sequence length",
        "size_in_GB": 0.16,
        "sources": {"hf": "jinaai/jina-embeddings-v2-base-de"},
        "model_file": "onnx/model_fp16.onnx",
    },

The quantized, onnx exported model is directly hosted by jinaai.

Changes

The following files were changed to add this model:

• fastembed/text/jina_onnx_embedding.py: Model definition was added

• tests/test_text_onnx_embeddings.py: Test was added, where the expected data was created with the supplied Colab-Notebook

  from sentence_transformers import SentenceTransformer
  model = SentenceTransformer('jinaai/jina-embeddings-v2-base-de', trust_remote_code=True)
  input_texts = [
      "hello world", "flag embedding"
  ]
  embeddings = model.encode(input_texts, normalize_embeddings=True)
  print(embeddings[0][:5])
  

  [-0.00857827  0.04176599  0.03420503  0.0309742  -0.01496792]
  

Tests

All tests passed after the changes.

[joein]

Hey @deichrenner,
thank you for the contribution!

I'll approve it as soon as the CI is green
It'll be available as of the next release (in the meantime you can use it from the main branch)