Grauho

Looking at rng.hpp and the de-noising code I believe this is already implemented correctly. Euler_a and dpm++2s_a get their random noise using the ggml_tensor_set_f32_randn function which sources it's rng from...

> On [this build](https://github.com/leejet/stable-diffusion.cpp/releases/tag/master-760cfaa) most of the lora's i have will cause a black image as an output. I'm not entirely sure why this would be the case, but the...

> I still cant reproduce the error. @grauho can you provide you setup? `./bin/sd --model ../../.models/sd1-5_PonyV6.safetensors --lora-model-dir ../../.loras/ --prompt " a lovely cat" ` CPU back-end without any extra compilation...

Updates: So for most LoRAs skipping the "alpha" scalars doesn't seem to matter. That said, when using certain LoRAs, such as the SDXL 8-Step Lightning LoRA mentioned above in gartia's...

https://github.com/fszontagh/sd.cpp.gui.wx This was posted in discussions the other day, haven't tried it yet but looks promising.

Looks like you've already tried to patch in some Vulkan back-end support, mind posting your fork?

I suspect this has to do with GGML currently only having support for up to 32-bit width integer and float types in it's ggml_type. It might be possible if one...

Could someone please link a model where they are having the F64 problem? I think I've put together a fix that at least seems to work with the LoRAs I...

I've written a small converter program in C that re-encodes entire safetensor files that does seem to do the job, once I put in handling for those using big-endian systems...

Alright here it is, feel free to give it a try. I was able to use it to convert both I64 and F64 containing models to something sdcpp could work...