Issues After Switching to the New Model Loading Logic
Hi,
First of all, thank you for creating and maintaining this workflow — it’s very well designed and extremely helpful. I really appreciate your work.
After switching to the new model loading logic, I encountered the following issues.
Used Diffusion Model Loader (KJ) to load:
ltx-2-19b-distilled-fp8_transformer_only.safetensors
Used Diffusion Model Selector to load:
ltx-2-19b-embeddings_connector_distill_bf16.safetensors
Both were connected as instructed.
Problem 1
With this setup, the CFG Guidance node throws:
'NoneType' object has no attribute 'model_options'
Problem 2
If I switch back to the previous UNet Loader:
The CFG error disappears.
But the Custom Sampler (Advanced) node throws:
'Linear' object has no attribute 'weight'
Could you please advise how this should be properly configured?
Thank you!❤
ComfyUI reverted some of the changes to allow people to have longer time to adapt to new ways.
Maybe something broke. Will take a look ;-)
Update:
Tried run the workflow my end, and no errors. (tried with the new way, and the old way )
Can you try this:
- Update ComfyUI to the very latests
- Update KJNodes to the very latest
Are the errors gone then?
I did update the I2V T2V Basic workflow though but mostly to remove a bit of clutter (NAG, chuck forward etc) that i'll leave for an "advanced" workflow.
To keep the basic one as simple as possible without too many elements.
(plus removed a prompt box left by accident from enhanced prompt workflow)