Google DeepMind 发布 740M 参数嵌入模型 EmbeddingGemma 2
Google 出了个只有 740M 参数的嵌入模型 EmbeddingGemma 2,能打过大它两倍的模型,Apache 2.0 开源,做端侧 RAG 搜索很合适。
Google DeepMind 发布 EmbeddingGemma 2,参数量仅 740M,但在多项基准上可比肩甚至超过两倍以上规模的专用模型。开发者可以用它为应用添加多模态搜索,例如通过语音备忘录定位视频中的片段。该模型还可与 Gemma 4 搭配,实现私有化的端侧 RAG。EmbeddingGemma 2 采用 Apache 2.0 许可证开源,权重已在 HuggingFace 和 Kaggle 上提供。
At 740M parameters, it’s competitive across benchmarks – even outperforming some specialist models more than twice its size. Developers can use it to add multimodal search to their apps – like finding moments in a video using a voice memo – or pair it with Gemma 4 for private, on-device RAG. EmbeddingGemma 2 is released under an Apache 2.0 license. Check out the weights on @HuggingFace and @Kaggle . Find out more → goo.gle/4y9wXgl 💬 1 🔄 0 ❤️ 13 👀 933 📊 3 ⚡
- Sundar Pichai10-06 16:03原文