KNOWLEDGE ATLAS (02513), in collaboration with Huawei, has open-sourced a new-generation image generation model named GLM-Image. This model was developed using the Ascend Atlas 800T A2 hardware and the MindSpore AI framework, completing the entire workflow from data processing to final training, making it the first SOTA multimodal model fully trained on domestic chips. GLM-Image utilizes an independently innovated hybrid architecture of "autoregression + diffusion decoder," achieving a union between image generation and language models, representing a significant exploration by KNOWLEDGE ATLAS into the new paradigm of "cognitive generation" technology, exemplified by devices like the Nano Banana Pro.
Comments