I had to swap out the GGUF Model Loader with this, to make it load the GGUF Text Encoder. https://github.com/molbal/ComfyUI-GGUF
I had to swap out the GGUF Model Loader with this, to make it load the GGUF Text Encoder.
https://github.com/molbal/ComfyUI-GGUF
I also had to edit:
\ComfyUI\custom_nodes\ComfyUI-MiniMaxH3-Cache_init_.py
line 176
from
slope_a = minimax_model.time_shift_slope(sigma_v, shift_v, shift_a).to(audio_out.dtype)
to
slope_a = minimax_model.time_shift_sigma(sigma_v, shift_v, shift_a).to(audio_out.dtype)
im confused? i didnt have that issue
im confused? i didnt have that issue
No issue here too - I loaded Qwen3vl-32B-MiniMax-H3-Q4_K_M into the ComfyUI CLIPLoader (GGUF), without any problem.
If by any chance your problem was that you were getting the error "Unexpected text model architecture type in GGUF file: 'qwen3vl'" at the CLIPLoader (GGUF) node, updating the node fixes it (going to the Extensions manager and updating ComfyUI-GGUF).