v4 with new wan 2.2? lora
means kijai or origional?
I used Kijai's version.
what is the base model ? also does it act like lora ... meaning fuse base wan2.2 with this ?
No, this is a "base model" that includes everything to get you started. Use it like shown on the "Model card".
Right so i need to get VAE , Text encoder etc from somewhere else to get it running in zerogpu
You cant possibly be serious, Rahul.. This is a full checkpoint. It has the model, clip and VAE built in, so for example in ComfyUI, you'd use the load checkpoint node instead of the load diffusion model node. The checkpoint node has outputs for Model, CLIP and VAE for this exact reason.
I am not using comfy.. i am using python script via wan2.2 pipeline https://github.com/Wan-Video/Wan2.2/blob/main/wan/configs/wan_t2v_A14B.py#L35 these info has to be provided to the pipeline . anyway will dig more
Idk if im too late and you found the solution, but im thinking you can probably find ways to just combine the functions of an LTX video code (for the ones that are full checkpoints like this one btw) with the Wan2.2 pipeline code for the actual loading of these things. Since, well, it's only loading we're talking about here.
Otherwise, I guess you could also just extract the CLIP, VAE and model from this checkpoint, and then use them as normal (or just the model, since you likely already have the CLIP and VAE)