[New Model]: VisionEncoderDecoderModel

[papayalove]

🚀 The feature, motivation and pitch

Does vllm have transformers VisionEncoderDecoderModel? We need to do image content recognition

Alternatives

No response

Additional context

No response

Before submitting a new issue...

• [X] Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[DarkLight1337]

~~No. Encoder-decoder architecture is currently limited to language-only models. We have plans to support it soon since multimodal Llama 3.1 will use it, see #8146 for more details.~~

[DarkLight1337]

Sorry, I seem to have misunderstood this model previously. It looks like this model is designed for generic multimodal encoder with generic language decoder, so it should be possible to implement in vLLM now. However, both encoder and decoder components need to be implemented in vLLM first, so it's not nearly as flexible as in transformers.

The closest thing to this right now in vLLM are the vision-language models that support arbitrary language decoder via init_vllm_registered_model.

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had any activity within 90 days. It will be automatically closed if no further activity occurs within 30 days. Leave a comment if you feel this issue should remain open. Thank you!

[github-actions[bot]]

This issue has been automatically closed due to inactivity. Please feel free to reopen if you feel it is still relevant. Thank you!