**AI 新锐袁进辉再创业，硅基流动获创新工场等 5000 万投资**
2023 年 2 月 23 日，OneFlow 创始人、光年之外联合创始人袁进辉宣布再次创业，成立新公司硅基流动（SiliconFlow)，专注做推理框架，服务大模型的应用。据《晚点 LatePost 》报道，硅基流动已完成 5000 万元人民币天使轮融资，由创新工场领投，耀途资本、奇绩创坛、美团联合创始人王慧文等跟投，投后估值为数亿元人民币。
袁进辉是人工智能领域的资深专家，拥有多年的大模型研发和应用经验。他曾担任百度深度学习研究院主任研究员，负责百度飞桨深度学习平台的研发工作。2018 年，袁进辉创立了 OneFlow，这是一个开源深度学习框架，旨在帮助开发者更轻松地构建和训练大模型。2021 年，袁进辉与其他几位合伙人创立了光年之外，这是一家专注于大模型应用的公司。
硅基流动是袁进辉的第三次创业项目。这家公司专注于推理框架的研发，旨在为大模型的应用提供更快的推理速度和更低的成本。推理框架是将训练好的大模型部署到实际应用中所必需的软件组件。它负责将大模型的模型参数加载到内存中，并根据输入数据对模型进行推理，从而得到预测结果。
硅基流动表示，其推理框架采用了创新的设计，可以显著提高大模型的推理速度和降低推理成本。该公司还表示，其推理框架将支持多种硬件平台，包括 CPU、GPU 和 FPGA。
硅基流动的成立，得到了业内人士的广泛关注。创新工场董事长李开复表示，硅基流动是一家非常有潜力的公司，其团队在人工智能领域拥有深厚的技术积累。他相信，硅基流动将能够在推理框架领域取得重大突破，并为大模型的应用带来新的机遇。
美团联合创始人王慧文表示，硅基流动是一家非常有前途的公司，其团队在人工智能领域拥有丰富的经验。他相信，硅基流动将能够在推理框架领域取得成功，并为大模型的应用带来新的可能。
硅基流动表示，其首款推理框架产品将于今年下半年发布。该公司还表示，其正在与多家企业合作，将推理框架应用于实际场景中。
英语如下：
**AI luminary Yuan Jinhui starts a new venture, SiliconFlow securestens of millions in funding for inference framework**
Keywords: Artificial intelligence, Entrepreneurship, Investment
**News Content: AI rising star Yuan Jinhui startsa new venture, SiliconFlow receives 50 million RMB in angel round funding from Innovation Works and others**
On February 23, 2023, Yuan Jinhui, founder of OneFlow and co-founder of BeyondCompute, announced that he is starting a new venture, SiliconFlow,a company dedicated to developing inference frameworks for large model applications. According to LatePost, SiliconFlow has completed a 50 million RMB angel round of financing led by Innovation Works, with participation from Yaotu Capital, Qiji Venture, and Meituan co-founder Wang Huiwen. The post-investment valuation is in the hundreds of millions of RMB.
Yuan Jinhui is a veteran in the field of artificial intelligence with years of experience in the research and development and application of large models. He previously served as the chief researcher at Baidu’s Deep Learning Research Institute, where he was responsible for the research and development ofBaidu’s PaddlePaddle deep learning platform. In 2018, Yuan Jinhui founded OneFlow, an open-source deep learning framework aimed at helping developers build and train large models more easily. In 2021, Yuan Jinhui co-founded BeyondCompute, a company focused on large model applications.
SiliconFlow is Yuan Jinhui’s third entrepreneurial venture. The company focuses on the research and development of inference frameworks, aiming to provide faster inference speeds and lower costs for large model applications. An inference framework is a software component necessary for deploying trained large models into practical applications. It is responsible for loading the model parameters of the large model into memory and performing inference on the model based on input data to obtain prediction results.
SiliconFlow claims that its inference framework employs an innovative design that can significantly improve the inference speed and reduce the inference cost of large models. The company also states that its inference framework will support multiple hardware platforms, including CPUs, GPUs, and FPGAs.
The establishment of SiliconFlow has attracted widespread attention from industry insiders. Kai-Fu Lee, Chairman of Innovation Works, said that SiliconFlow is a company with great potential, and its team has a deep accumulation of technology in the field of artificialintelligence. He believes that SiliconFlow will be able to make significant breakthroughs in the field of inference frameworks and bring new opportunities for the application of large models.
Wang Huiwen, co-founder of Meituan, said that SiliconFlow is a very promising company with a team that has extensive experience in the field of artificial intelligence. He believes that SiliconFlow will be able to succeed in the field of inference frameworks and bring new possibilities for the application of large models.
SiliconFlow said that its first inference framework product will be released in the second half of this year. The company also said that it is working with multiple companies to apply the inference framework in real-world scenarios.
【来源】https://www.latepost.com/news/dj_detail?id=2074
Views: 1