1.3B Model GGUF?
Hi Calcuis
Thank you very much for your work. The LTX-GGUF r2 are much better then before, and really good now. With the WAN-Model, I had a lot of struggle with the your GGUF-Version. But now after a mix of versions, it is working. But my hardware is to weak for the big model, the small WAN Model T2V is working really great. Normally there is no need for an gguf-version, and perhaps, it makes no sense. But I want to ask anyway: Is there a plan for a gguf version of the t2v-1.3 model? My hope is, that if the model is more tiny, then I could increase the size of the image. But to be honest, I do not know, if that is true.
1.3b seems a test model; not really ready for production; the tensors were trimmed up already; you could try to convert it with convertor zero and see does it work or not first
ok. I tested the zero convertor. It is really as simple as it could be. it is fast also. The following video creation gguf-run then was finished with an error, and the gguf-file was slightly larger then the safetensor :-). But this converter is really a good thing, when the returning file is working. I was surprised that I could not define the bit level of quantification. But it has chosen by itself. Anyway, thank you for your work.
it works for most of the common models, including encoder and decoders; but not all, since some of them have unique structure and compatibility problem, might need extra editing work
you are lucky; finally make 1.3b gguf works; uploading
Oha, you are so incredible diligent. If I see the model, I will download it and after testing it, I will report. Thank You.
Hi Calcuis
First Feedback: Original safetensor and your gguf were equal in picture quality, in generation speed and memory consumption, too. I am surprised that the size of the gguf is nearly the double . But I need to do further tests to decide if there is any special advantage of the gguf. Thank you.
Another question: I see you have so much stuff like pig. Picture generation and Video generation, too. I have lost the overview of all the possibilities. Is there somewhere a place, where I can inform myself, why and how I should use this or that?
Thank you for your work!!