shiyu (@shiyu17)哪个嵌入模型识别pdf,图片精准 中发帖

部署了一个ollama,想弄个嵌入模型识别pdf形成知识库,再通过deepseek对话引用知识库内容,试了chroma/all-minilm-l6-v2-f32:latest和nomic-embed-text:latest感觉都不太精准