Ean Garvey

@gpetters94 could this be related to the issue you're having with multicontrolnet? This sounds a lot like how you described the behavior.

Right, we have an implementation in the works to enable StableDiffusionXL in SHARK, but until https://github.com/nod-ai/SHARK/pull/1731 is landed any of these `XL` models from HF/diffusers/CivitAI will see this same issue.

Merged and reverted unfortunately. It's still in progress! @Abhishek-Varma will know best what's left to be done for SDXL support.

It's on the way https://github.com/nod-ai/SHARK/pull/1941 custom SDXL models is still up to a week or two out... it's quite a chunky model and the UX needs quite a bit of...

This is going to be lost in the switch to SHARK-2.0 as it targets the old app code. We should add this as a follow-up to the beta release. If...

https://github.com/nod-ai/SHARK-Turbine/issues/152

Thanks for the heads up. I'm looking into this right now, trying to reproduce. If there are any patterns in which unet/vae/lora combinations I should look at, let me know.

> @monorimet The specific combination used in this example was as follows: > > > > Model: [Lily Mix v2](https://civitai.com/api/download/models/186820?type=Model&format=SafeTensor&size=pruned&fp=fp16) > > VAE: [kl-f8-animev2](https://civitai.com/api/download/models/189192?type=VAE&format=Other) > > LoRA: [Secret Sauce -...

Something changes in the unet model when we use LoRAs. It seems to work fine without the tunings, but with tunings I reproduce your error.  I'll hypothesize that the...

Hey, it might help to use the new `--device_allocator_heap_key` option when launching the .exe. if you have a .bat file wrapping the webui launch, or if invoking from the command...