City

>昨天 I've pushed changes ~30 minutes ago ( https://github.com/city96/ComfyUI_ExtraModels/commit/5101719ca859a2d75f5e830230011668213b559d ) that might fix the issue, so please update again. If you see `loading in lowvram mode` in the console then...

You can use the default ComfyUI load clip node with the SD3 fp8 weights + the pixart T5 from SD3 clip node. Here's a workflow for it: [PixArtV4.json](https://github.com/user-attachments/files/16400437/PixArtV4.json) 

Hi. If you don't mind, could you share one of the LoRA files so I can implement it? The current implementation is largely based on the peft LoRA from the...

No clue if that fixes it but I made it match the correct keys. It still has a bunch of "missing keys" listed but looks like OneTrained just doesn't train...

The amount of convoluted jank and technical debt in this repo is staggering, and I should really just rewrite the entire LoRA loading logic instead of having half the key...

I've had a play around with Lumia as well as [Large-DiT](https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/Large-DiT-T2I) in the past, but they were pretty undertrained. Looks like it got updated recently and no longer uses llama...

To me that looks like the git clone created a subfolder in the custom nodes folder (`custom_nodes\\custom_nodes`) instead of cloning it normally. The command from the readme assumes you're just...

Yeah, this definitely seems odd, I can't really think of anything that would cause it. Could you maybe tell me the following: - What model file(s) you used - Python...

Well, no issues with the workflow or the pixart model file, so it must be something else. I assume there's no errors/warnings in the console? (you can ignore the "Not...

Sana team uploaded a copy of the gemma2 weights that don't need an access token to download, it should be available in the dropdown now. https://github.com/city96/ComfyUI_ExtraModels/commit/4a770ac22b85e70de5bafd5f6b68118dddcef2fd