近日，阿里云宣布开源其大规模人工智能模型通义千问，包括720亿参数和18亿参数模型，以及音频大模型Qwen-Audio。这些模型已在阿里魔搭社区上线，为广大开发者提供免费使用。
通义千问的720亿参数模型（Qwen-72B）和18亿参数模型（Qwen-1.8B）分别代表了阿里云在自然语言处理和音频处理领域的最新成果。除了预训练模型，阿里云还同步推出了对应的对话模型，便于开发者们进行推理训练。
为了让开发者更容易上手，阿里云还提供了72B和1.8B对话模型的4bit/8bit量化版模型。这些量化版模型在小样本条件下表现出色，能有效提升开发效率。
此次开源举措体现了我国在人工智能领域的发展势头，将推动国内外开发者共同创新，为各行各业提供智能解决方案。
英文翻译：
News Title: Alibaba Cloud Opensources Large-scale AI Models
Keywords: Alibaba Cloud, Open-source, Large-scale AI Models
News Content:
Recently, Alibaba Cloud announced the open-source of its large-scale AI model Tongyi Qianwen, including 7.2 billion parameter and 180 million parameter models, as well as the audio grand model Qwen-Audio. These models have been launched on the Alibaba Moda community, providing developers with free use.
Tongyi Qianwen’s 7.2 billion parameter model (Qwen-72B) and 180 million parameter model (Qwen-1.8B) represent Alibaba Cloud’s latest achievements in the fields of natural language processing and audio processing. In addition to the pre-trained models, Alibaba Cloud also launched corresponding dialogue models for developers to conduct reasoning training.
To make it easier for developers to get started, Alibaba Cloud provides 4bit/8bit quantized versions of the 72B and 1.8B dialogue models. These quantized versions perform well under small sample conditions, effectively improving development efficiency.
This open-source initiative reflects the development momentum of China’s AI field and will promote innovation among domestic and foreign developers, providing intelligent solutions for various industries.
【来源】https://www.modelscope.cn/headlines/article/269
Views: 1