search

AUTOMATIC1111 / stable-diffusion-webui-tensorrt

MIT License
311 stars 20 forks source link

What is meaning "Loras are baked in into the converted model. "? #40

Open edisonchan opened 1 year ago

edisonchan commented 1 year ago

Does it mean if want LORA works we need to merge the LORA into checkpoint model before convert to onnx ?

jebarpg commented 1 year ago

You need to go into settings>extra networks and find the textbox for loras to add all the loras you want to be added into the TRT model when you convert.

edisonchan commented 1 year ago

You need to go into settings>extra networks and find the textbox for loras to add all the loras you want to be added into the TRT model when you convert.

I did what you said to convert to a onnx file and then convert to a .trt file, but the trigger word of lora still not work when create image via text to image.

jebarpg commented 1 year ago

hmm... okay. I'm not sure then. I've got very limited knowledge of TensorRT. I've been trying to get answers about how they calculated the size of the shape on the NVIDIA repo but have yet to get a response. I would say that at this point in time you might just go with merging the LORA into the checkpoint then converting it over since it isn't working with the Extra Networks.