Open jasd3199 opened 1 week ago
Hi, could you provide the inference script and model type?
More information would help us quickly pinpoint the issue.
Yes, I'm using ComfyUI for the inference with one of the example workflows that the repo provides (hunyuan_ksampler_api.json), only changing the desired size and prompts.
The model that I loaded was in: https://huggingface.co/Tencent-Hunyuan/HunyuanDiT/blob/main/t2i/model/pytorch_model_ema.pt
I had this problem using the initial version and the last version V1.1
we will fix it and release code around 6.30.
When I change the desired size to any different to width=height, the generation fails.
This is an example (1280x720):![hunyuan_00034_](https://github.com/Tencent/HunyuanDiT/assets/80776713/612d9782-15ee-4653-9d7e-dd8451cb151f)