jayn7

/

WAN2.2-I2V_A14B-DISTILL-LIGHTX2V-4STEP-GGUF

like 36

Done! lmk if there are any issues.

@jayn7 Thank you!,one issue though, I noticed that the link is the full precision model 😅, which runs slower on my setup apparentlly. Would it be possible to create a GGUF version from this alternative source? https://huggingface.co/obsxrver/wan2.2-i2v-lightx2v-260412/tree/main
it's already fp8 scaled but the author said he might convert it to FP16/BF16 from the full precision, I’ll let you know if he already convert it. Thanks, and sorry for the bother.

@Fronyax did you test it by yourself?

Every GGUF I upload here always starts from half precision BF16/FP16. So yes, I already convert/downcast the FP32 model to BF16 before quantizing it to GGUF.

@jayn7 Oh, good to know. That means your 260412 GGUF is already perfect then, no issues.
Yes I tested your 260412 GGUF against your 1030 GGUF, and the 1030 is a little bit faster.