Add support for embedding input

[Andcircle]

I saw similar ask for this feature, but seems like still open-ended, not solved yet.
We're using different model e.g. SAM / SigLIP as LEGO component to build our own VLM, only language model is time consuming.
Hope we can prepare embedding input, and feed it into language model with VLLM.

Just trying to confirm, do we already have this feature? If yes, could somebody share an example code.
If not, is this feature on roadmap?

Really appreciated.

[yuanzhiyong1999]

Have you found an alternative?

[Andcircle]

Nope, keep waiting 😅

…

On Thu, Jul 25, 2024 at 9:02 PM Zhiyong Yuan ***@***.***> wrote: Have you found an alternative? — Reply to this email directly, view it on GitHub <#6316 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AHYYPBIRSM4JOP43T4RQH6DZOHC4HAVCNFSM6AAAAABKV2ONM6VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTAMJVGUZDIMQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>