Maybe something wrong with custom model

[Ar57m]

it appears that the app wants 2 tokenizer_config.json in the same folder, which is impossible. On presentation/src/main/java/com/shifthackz/aisdv1/presentation/screen/setup/ServerSetupScreen.kt line 618 do you mean vocab.json?

Another question, I converted to ort successfully this, right here do I use the model.with_runtime_opt.ort or the normal one? and does the app needs the model_index.json?

[Ar57m]

another thing, is the folder vae_encoder not needed?

[ShiftHackZ]

The correct folder structure should be:

Regarding your oher questions:

do you mean vocab.json?

Yes, vocab.json shoud be there.

do I use the model.with_runtime_opt.ort or the normal one?

I used a regular model.ort in my experiments.

and does the app needs the model_index.json?

another thing, is the folder vae_encoder not needed?

Current implementation does not use model_index.json and vae_encoder, but it may change in future implementation.

[Ar57m]

Another question, I converted to ort successfully this, right here

Thanks for replying, tried running it but got this error, it's a sd turbo model.

tried this too but got other error(forgot to screenshot)

[Ar57m]

here sdturbo I converted to onnx, but got an error while converting it to ort

Error converting /content/sdturbo/unet/model.onnx: [ONNXRuntimeError] : 1 : FAIL : inference_session.cc:806 SaveToOrtFormat Failed to save ORT format model to file: /content/converted/unet/model.ort Converting