What is minimax_h3_lynnreal_light_vae_int8_convrot?
Is it like an even more optimised video VAE?
It's this one, converted and quanted to int8: https://huggingface.co/stdstu123/LynnReal-Onmi-light-vae
It's about 1.3x faster for decoding.
It's this one, converted and quanted to int8: https://huggingface.co/stdstu123/LynnReal-Onmi-light-vae
It's about 1.3x faster for decoding.
Do I need the custom node mentioned there? With native it's giving error. Thanks for the reply.
It's this one, converted and quanted to int8: https://huggingface.co/stdstu123/LynnReal-Onmi-light-vae
It's about 1.3x faster for decoding.
Do I need the custom node mentioned there? With native it's giving error. Thanks for the reply.
No, but you need a nightly ComfyUI build currently as the support was just added yesterday.
So - it is slower than older in8_convrot (2x vs 1.3x speed up). Aye?
So - it is slower than older in8_convrot (2x vs 1.3x speed up). Aye?
Nope, it's 1.3x on top of that.
So - it is slower than older in8_convrot (2x vs 1.3x speed up). Aye?
Nope, it's 1.3x on top of that.
Whooaa! That's some good news!
Cheeers!
Is it free speed? or...
I don't see benchmarks. They only say the colors would change.