TencentAnnounced its Hunyuan WenshengtuLarge ModelUpgrade and openOpen Source, which has been released on Hugging Face and Github, contains a complete model including model weights, inference code, model algorithms, etc., which can be used by enterprises and individual developersFree for commercial use.
▲ Hunyuan Wenshengtu effect
The upgraded Hunyuan Wenshengtu model uses the same DiT architecture as Sora. Tencent said that Hunyuan DiT isThe first bilingual DiT framework in Chinese and EnglishHunyuan DiT is a text-to-image generation model based on the Diffusion transformer. This model has the ability to understand Chinese and English in fine-grained terms. Hunyuan DiT can have multiple rounds of dialogue with users and generate and improve images based on the context. This is also the first in the industry.The first native ChineseThe DiT architecture Wenshengtu open source model supports bilingual input and understanding in Chinese and English, with 1.5 billion parameters.
Running this model requiresCUDA-enabled NVIDIA GPU, required to run Hunyuan DiT aloneMinimum video memory is 11GB, running DialogGen (a text-to-image multimodal interactive dialogue system launched by Tencent) and Hunyuan DiT at the same timeAt least 32GB of video memory is required, Tencent said they have tested Nvidia's V100 and A100 GPUs on Linux.
▲ Configuration requirements
According to previous reports, the results of the first official "large model standard conformity assessment" in China were announced. Tencent's Hunyuan large model became the first batch of domestic large models to pass the assessment. Other large models that passed the assessment included Alibaba Tongyi Qianwen, 360 Zhinao and Baidu Wenxin Yiyan.