LTX2 comfyui native WF, Error running sage attention.

[m8rr]

#399
#402

I applied PR and it worked fine, except sageattention.
Error running sage attention: Input tensors must be in dtype of torch.float16 or torch.bfloat16, using pytorch attention instead.

[zwukong]

true

[MeiYi-dev]

#399 #402

I applied PR and it worked fine, except sageattention. Error running sage attention: Input tensors must be in dtype of torch.float16 or torch.bfloat16, using pytorch attention instead.

remove the --use-sage-attention arg from the startup args and use the patch sage attention node from KJNodes