HomeAI News
Alibaba Cloud open source Tongyi Qianwen 7 billion parameter model, free and commercially available

Alibaba Cloud open source Tongyi Qianwen 7 billion parameter model, free and commercially available

Hayo News
Hayo News
August 3rd, 2023
View OriginalTranslated by Google

On August 3, ModelScope, an AI model community, launched two open source models, Qwen-7B and Qwen-7B-Chat. Alibaba Cloud confirmed that it is a general model and a dialogue model with 7 billion parameters of Tongyi Qianwen. Both models are open source, Free and commercially available. In multiple authoritative evaluations, the Tongyi Qianwen 7B model has achieved far more effects than models of the same size at home and abroad, and has become the strongest Chinese and English 7B open source model in the industry.

In April this year, Alibaba Cloud launched a self-developed large-scale model Tongyi Qianwen, which triggered a blowout market demand. The open source version of the miniaturized model is expected to smooth the threshold for using the model, allowing a large number of small and medium-sized enterprises and AI developers to use Tongyi Qianwen earlier and faster. This move also made Alibaba Cloud the first large-scale technology company in China to join the ranks of large-scale open source.

Qwen-7B is a base model that supports Chinese, English and other languages. It is trained on more than 2 trillion token data sets, and the length of the context window reaches 8k. Qwen-7B-Chat is a Chinese-English dialogue model based on the pedestal model, which has achieved alignment with human cognition. The open source code supports the quantization of Qwen-7B and Qwen-7B-Chat, allowing users to deploy and run models on consumer-grade graphics cards.

Users can download the model directly from the Mota community, or access and call Qwen-7B and Qwen-7B-Chat through the Alibaba Cloud Lingji platform. Alibaba Cloud provides users with model training, reasoning, deployment, fine-tuning, etc. full range of services.

Tongyi Qianwen 7B pre-training model has performed well in multiple authoritative benchmark evaluations. Its Chinese and English capabilities far exceed those of open-source models of the same scale at home and abroad, and some capabilities even surpass 12B and 13B open-source models.

On the English proficiency assessment benchmark MMLU, the Tongyi Qianwen 7B model scored higher than other 7B, 12B, and 13B mainstream open source models. The benchmark includes English questions in 57 subjects, testing the comprehensive knowledge and problem-solving ability in the fields of humanities, social sciences, science and engineering.

On the C-Eval benchmark of Chinese general knowledge ability, Tongyi Qianwen is the 7B open source model with the highest score in both the verification set and the test set, showing a solid Chinese ability. Compared with the lively AI open source ecology in the English-speaking world, the Chinese community lacks excellent base models. The addition of Tongyi Qianwen is expected to provide more choices for the open source community and promote the construction of China's AI open source ecosystem.

On benchmarks such as the mathematics problem-solving ability evaluation GSM8K and the code ability evaluation HumanEval, the Tongyi Qianwen 7B model also performed well, outperforming all open source models of the same size and some large-scale open source models.

Alibaba Cloud said that the open source large model can help users simplify the process of model training and deployment. Users do not have to train the model from scratch, but only need to download the pre-trained model and fine-tune it to quickly build a high-quality model.

Reprinted from 央广网View Original