Stopped working with int8 models after ComfyUI update when using dynamic-vram

[alecz20]

When using int8_convrot Diffusion model and clip with "--disable-smart-memory", unloading the clip fails.

I simply updated ComfyUI and its dependencies and re-runing the exact same workflow from before the update produces:

[INFO] Requested to load Krea2TEModel_
[INFO] Model Krea2TEModel_ prepared for dynamic VRAM loading. 4604MB Staged. 0 patches attached. Force pre-loaded 249 weights: 627 KB.
[ERROR] !!! Exception during processing !!! No backend can handle 'dequantize_int8_embedding': eager: q: dtype torch.bfloat16 not in {torch.int8}
[ERROR] Traceback (most recent call last):
...
ComfyUI/venv/lib/python3.12/site-packages/comfy_kitchen/registry.py", line 233, in get_capable_backend
    raise NoCapableBackendError(func_name, failures)
comfy_kitchen.exceptions.NoCapableBackendError: No backend can handle 'dequantize_int8_embedding': eager: q: dtype torch.bfloat16 not in {torch.int8}

This doesn't happen with the regular native CLIP Text Encode node. (though the result is not as good obviously)