[Feature Request] Support the standard Lora embedding format <lora:_>

[nvidia]

[Feature Request] Support the standard Lora embedding format

from stable-diffusion-webui-tensorrt.

[contentis]

I'd like to get this working in the future, but I cannot give you an ETA on when this will be possible. As this is a side-project, I need to find time, as this will need work on core components. And especially in these first days it's gonna be bug fixing.
Stay tuned for updates.

[MorkTheOrk]

Hey! This is not easily possible in this extension due to the access we get from automatic1111s webui, we provide a function that runs the Unet inference and is being called by the webui, we dont have control over prompt parsing. Additionally, there is no current way for us to refit the engine with the LoRA directly from the checkpoint.

[ArrowM]

Hm, I know extensions can read and modify prompts prior to inference, for example https://github.com/adieyal/sd-dynamic-prompts. Would it not be possible to read the prompt prior to inference, check for a matching Lora TRT using the hash and then use that instead of the input from the dropdown.