Sakana AI，一家由知名人工智能专家Lion Jones和前Stability AI研究主管David Ha共同创立的大模型初创公司，近日宣布推出一种名为“进化模型合并（Evolutionary Model Merge）”的革命性技术。这一技术受到自然界进化和自然选择的启发，旨在自动发现如何最佳地结合不同的开源模型。
Sakana AI同时发布了三个基于这一新技术的日语生成模型：EvoLLM-JP、EvoVLM-JP和EvoSDXL-JP。这些模型已经在GitHub和Hugging Face平台上发布，向开发者社区展示了模型合并技术的巨大潜力。
EvoLLM-JP是一款日语语言模型，能够生成流畅且具有高度语境相关性的文本。EvoVLM-JP则是一款结合了语言理解和图像生成能力的视觉语言模型，而EvoSDXL-JP则是一款专注于图像生成的大模型，能够创造出高分辨率的原创图像。
Sakana AI的这一突破性进展，标志着人工智能模型开发领域迈入了一个新的时代。通过模型合并技术，开发者将能够更有效地利用现有的开源模型，创造出更加强大和多样化的AI应用。
这一消息在人工智能和科技界引起了广泛关注，预示着未来人工智能模型的发展将更加智能化、高效化。随着这些新技术的不断迭代和完善，AI在多个领域的应用将迎来新的飞跃。
英文翻译内容：
Title: Sakana AI Introduces Evolutionary Model Merging Approach, Releases Three Japanese Generative Models
Keywords: Model Merging, AI, Evolutionary Algorithms
News content:
Sakana AI, a startup specializing in large-model development co-founded by Lion Jones, one of the authors of the seminal paper “Attention Is All You Need,” and former research director at Stability AI, David Ha, has unveiled a groundbreaking technique called “Evolutionary Model Merge.” Inspired by the principles of evolution and natural selection, this technique automates the process of discovering how to combine various open-source models in the most effective manner.
Sakana AI has also released three Japanese-language generative models based on this new technology: EvoLLM-JP, EvoVLM-JP, and EvoSDXL-JP. These models are now available on GitHub and Hugging Face, showcasing the potential of model merging to the developer community.
EvoLLM-JP is a language model capable of generating text that is both fluent and contextually relevant. EvoVLM-JP combines language understanding with image generation capabilities, while EvoSDXL-JP is focused on image generation, producing high-resolution original images.
This breakthrough by Sakana AI marks a new era in the field of AI model development. With the model merging technique, developers will be able to leverage existing open-source models more effectively, creating a wider array of powerful and diverse AI applications.
This news has generated significant attention within the AI and tech communities, signaling the advent of a new era in AI model development, characterized by increased intelligence and efficiency. As these new technologies continue to evolve and improve, AI applications across multiple fields are expected to experience significant advancements.
【来源】https://sakana.ai/evolutionary-model-merge/
Views: 1