Here is my quantized FP8 Version of the Wan2.1_14B 720p t2v model so that we can run it on 50 and 40 series or for even faster inference on bigger cards.
Description
FAQ
Comments (8)
Would you be able to do a quantized FP8 version of T2V-1.3B for those of us with lower end hardware?
yes will do
WackyLabsFeb 28, 2025
Awesome, thanks!
@WackyLabs here you go https://civitai.com/models/1307708/wan21t2v13bfp8?modelVersionId=1475726
@yogotatara Thank you!
AMark78Mar 24, 2025
Wow! I think what we really need is quantized I2V-1.3B. Because even on a 4090 a generation is taking about 40mins. Is that even possible?