alexdsh
alexdsh
so it's not the processor that's overloaded, but the graphics core., regarding the model, I use gemma2-2B q4fp16.mlc, it's already quantized to the maximum, besides, I also launched gemma2-7B-int1.gguf (though...
realsr bigimage?
file in archive with model for win x64 https://github.com/vitoplantamura/OnnxStream/releases/download/v0.1/StableDiffusion-OnnxStream-Windows-x64-with-weights.rar
I understand you, I'll try and write about the result
everything works, it counts steps, but everything is insanely slow. one step takes 673843 ms (almost 12 minutes?). then none of the keys from the android version work, tried to...
Thanks, Vito! I also wanted to clarify the --step parameter, does it work or not? I mean, is it possible to change the number of steps to an arbitrary one?
Hello, Vito! I tried all variations of commands and launching directly from the models folder. Only the default model (sd 1.5) works, which comes with sd.exe. The --rpi key, which...
Hello! It works, the turbo definitely generates output, but the SDXL starts and runs slowly. I selected step 4. But I still haven't seen any results after more than three...
Hello! In the question I named the model sdxl 1.0 without the turbo prefix. As far as I understood from the author's description, it is sdxl 1.0 that is trained...