Pangu 3.0

Se publicó en July 2023. La capacidad por parámetro ha mejorado considerablemente desde entonces, por lo que un modelo más reciente del mismo tamaño suele ser el mejor uso del mismo hardware.

Funciona en el dominio de Multimodal, Language, Image generation, Vision, y se registra como manejando la tarea de language modeling/generation, Image generation. Un modelo puede manejar varios de cada uno, así que estas son las áreas para las que fue construido en lugar de un límite en lo que intentará.

Ninguno. Este es un modelo cerrado: sus PESOS nunca fueron publicados, por lo que no se puede descargar o ejecutar en su propio hardware a ningún precio. Solo se puede acceder a él a través de su proveedor.

No. Sus pesos no han sido publicados, por lo que existe solo como un servicio controlado por su propietario.

Tiene un recuento de parámetros de 100B. 100B? I think the five foundation models are all included in the same system, instead of being five different variants of Pangu, but that's not very clear. I think that's implied by "All of these capability sets will be made available to customers and partners, and will be consistent regardless of the model size". https://www.huaweicloud.com/intl/en-us/news/20230707180809498.html "Huawei Cloud Pangu models were designed to focus on the practical needs of specific industry scenarios. The newly-launched Pangu Models 3.0 use a "5+N+X" three-layer architecture. The L0 layer consists of five foundation models: NLP, CV, multimodal, prediction, and scientific computing, which provide general skills to power an endless possibility of industry-specific applications. Pangu Models 3.0 are available in different sizes: 10 billion parameters, 38 billion parameters, 71 billion parameters, and 100 billion parameters, meeting diverse customer needs and different standards on latency and response times. Brand-new capability sets are also provided, such as knowledge-based Q&A, copywriting, and code generation for the Pangu NLP model; and image generation and understanding for the Pangu multimodal model. All of these capability sets will be made available to customers and partners, and will be consistent regardless of the model size (number of parameters). The L1 layer consists of N industry-tailored models. Huawei Cloud can provide customers with industry models it has trained on open industry datasets, including Pangu models for government, finance, manufacturing, mining, and meteorology. Alternatively, customers can train their own models using their own datasets based on Huawei's L0 or L1 Pangu models. The L2 layer provides pre-trained models for specific industry scenarios and tasks, such as intelligent government hotline, intelligent branch assistant, lead compound screening, conveyor belt foreign object detection, and typhoon trajectory prediction. These models can be quickly deployed off-the-shelf.". Esa cifra es el total, y es lo que decide cuánta MEMORIA necesita el modelo — aproximadamente medio gigabyte por mil millones en la compresión que la mayoría de las personas usa.

Fue publicado por Huawei, basado en China, una organización categorizada como industry.