As i stated on reddit, great job on the doing something marvelous. , i have tested it on aitoolkit, and musubi-tuner. Now for my question how do you do wan2.2 I2v? It seems both will only do t2v which work very well. Or maybe i'm missing something simple? I've upgraded comfyui to the latest, but it made no difference. I used the same exact fp16 as the t2v, but of course just did the i2v version, both for low noise.
musubi-tuner throws this error when trying wan2.2 i2v-[musubi-tuner] RuntimeError: Error(s) in loading state_dict for WanModel:
[musubi-tuner] size mismatch for patch_embedding.weight: copying a param with shape torch.Size([5120, 36, 1, 2, 2]) from checkpoint, the shape in current model is torch.Size([5120, 16, 1, 2, 2]).
which i wonder is if it's because it's flagging i2v as false?
INFO:musubi_tuner.wan.modules.model:Creating WanModel. I2V: False,
And as far as aitoolkit we don't enter which model we use.
yes i tried the t2v file in a i2v environment and it just does not work. It's not cross compatible like the wan2.1 loras. Put the lora in t2v and it does great at .75 strength for the loras it made. I started with just training in low. Then i tried training both, still i2v a no go with the t2v loras made.
Changing anything inside the json has no effect on the dif model picked. Changing the example model in aitoolkit doesn't do it. I do have the diffusion models for both programs in the proper spot. I was even looking for a clue in your realtime lora folder in comfyui. I tried changing the template there with no luck. Any help would be appreciated. Thank you.
rig spec-suprim 5090, 64gb ram
As i stated on reddit, great job on the doing something marvelous. , i have tested it on aitoolkit, and musubi-tuner. Now for my question how do you do wan2.2 I2v? It seems both will only do t2v which work very well. Or maybe i'm missing something simple? I've upgraded comfyui to the latest, but it made no difference. I used the same exact fp16 as the t2v, but of course just did the i2v version, both for low noise.
musubi-tuner throws this error when trying wan2.2 i2v-[musubi-tuner] RuntimeError: Error(s) in loading state_dict for WanModel:
[musubi-tuner] size mismatch for patch_embedding.weight: copying a param with shape torch.Size([5120, 36, 1, 2, 2]) from checkpoint, the shape in current model is torch.Size([5120, 16, 1, 2, 2]).
which i wonder is if it's because it's flagging i2v as false?
INFO:musubi_tuner.wan.modules.model:Creating WanModel. I2V: False,
And as far as aitoolkit we don't enter which model we use.
yes i tried the t2v file in a i2v environment and it just does not work. It's not cross compatible like the wan2.1 loras. Put the lora in t2v and it does great at .75 strength for the loras it made. I started with just training in low. Then i tried training both, still i2v a no go with the t2v loras made.
Changing anything inside the json has no effect on the dif model picked. Changing the example model in aitoolkit doesn't do it. I do have the diffusion models for both programs in the proper spot. I was even looking for a clue in your realtime lora folder in comfyui. I tried changing the template there with no luck. Any help would be appreciated. Thank you.
rig spec-suprim 5090, 64gb ram