About gguf version 0.9.1
I believe there might be an issue with GGUF 0.9.1.
This GGUF gives the same results as GGUF 0.9.0.
For example, if I compare GGUF 0.9.1 and GGUF 0.9.0, the results are almost 97% identical.
However, if I compare the original non-GGUF model 0.9.1 and 0.9.0, the results are completely different.
Could there be an issue with GGUF 0.9.1 Q8 and Q4?
Something is not right.
Here a quick workflow for test and see the differences.
https://www.mediafire.com/file/m4atam4pmrvpzw5/test_ltx_gguf.json
Results:
GGUF 0.9.0
GGUF 0.9.1
Original 0.9.0
Original 0.9.1
did you fix the seed (under sampler)? what is your prompt? ok, seems the same structure
Yes I have give you a simple workflow, you can see the difference.
thanks; according to the console statistics, the new set is faster than before; will look into the source file, see anything mess up
If you can't test my workflow here a quick video exemple.
https://www.youtube.com/watch?v=jCnH3kl1Sic
thanks for the video review and output comparison; just re-converted the original safetensors, the recent general.file_type is [1], but it should be [32]; which is not supported by the node recently; should wait the code update first then re-quantize those files again
Good, thank you!
If any users are reading this message, I’ve made some updates to my workflow ltx model test.
This workflow tests both models 0.9.0 and 0.9.1, includes an experimental I2V method, and provides an example demonstrating how to extend the T2V mode video.
If it’s helpful to you, that makes me happy.
https://www.mediafire.com/file/dp04l3oi884054k/Dave_Gravel_LTX_TEST_PLUS_LONG_VIDEO_EXEMPLE_V2.zip