Huawei
Huawei denies allegation of training Pangu AI model on rival company’s data
Looks like the recently open-sourced Huawei Pangu AI model is slammed with allegation for being trained on other companies’ data. Although the tech giant refutes these claims and responds to this matter with sincerity.
On June 30, Huawei made two of its Pangu AI models available for all developers. The company announced an open-source strategy for the Pangu 7B dense model and the Pangu Pro MoE 72B hybrid expert model.
This strategy will enable developers to use the original codes of the Pangu AI models and modify them as per their app or code-making requirements.
However, the newly open-sourced Huawei Pangu AI model (72B hybrid expert) faced allegation from Alibaba Cloud. The rival firm says that the company’s AI model has been trained on its Qianwen Qwen-2.5 14B model, and accuses it of copyright issues.
Today, Huawei has responded to this matter, saying that though the Pangu 72B model integrates certain codes of other large open-source models, it is not trained on any other company’s data. Further, Huawei said that the model is based on Ascend chips and uses the same for development and training.
(Image Credits: Huawei)
Huawei reveals that the company has made the Pangu Pro MoE 72B model with top-notch components. It has also expanded its scope of innovation and used a significant architectural design as well as technical features for the new model.
The Chinese firm has trained Pangu Pro MoE model on Ascend GPUs and NPUs. It is capable of solving the load balancing problem of large-scale distributed training and also enhances the overall training efficiency. You can check more details about the new model below.
(source)
