Question about using the cfg_scale (Guidance) parameter with HunyuanImage‑3.0‑Instruct‑Distil
Hi everyone,
I’m currently working with the tencent/HunyuanImage‑3.0‑Instruct‑Distil model, and I would like to know whether it supports a cfg_scale (Classifier‑Free Guidance) parameter during image generation.
The documentation does not explicitly mention a guidance scale or an equivalent setting, unlike what we typically see in models such as Stable Diffusion or Flux. Before modifying the pipeline or adding a custom wrapper, I’d like to clarify a few points:
Does this model support any form of guidance or prompt‑strength parameter?
If yes, what is the exact parameter name and the recommended value range?
If not, is this a limitation specific to the distilled version, or to the Hugging Face pipeline implementation?
Is there an official or recommended way to adjust prompt adherence or instruction strength?
Thank you in advance for your help, and congratulations to the team for the impressive work on this model.