Release Plans for FP8 models? GPU Minimum Requirement?
Hi @TencentOpen @TencentAIGC-Lab , I have some questions about your release plans:
Would you have any plans to release FP8 model for this
HunyuanVideo-PromptRewritefine-tuned model, like you have already done with the Hunyuan-A52B-Instruct-FP8?Any release plans for GGUF or any other smaller/faster quantized models?
What is the minimum GPU requirement for inference of this model? Like 16x H100 GPUs? Could we use A100 80GB VRAM GPUs instead?
Any cloud/API services (Tencent or anywhere else) that we can immediately use this PromptRewrite model, instead of self-hosting?
Is this PromptRewrite model still compatible with the newly released I2V model? If not, any release plan for I2V versions?
I'm really eager to try it to see how this rewrite model would improve the Hunyuan Video T2V and I2V outputs, but the model is prohibitory huge for me or most of the open-source community.
This PromptRewrite model should be the greatest competitive advantage over other open-sourced video generation models, such as Wan2.1, if this PromptRewrite model is more accessible to us.