📌 📄 内容摘要:
vfx (@freebsdfx) 在 测试了一下本地翻译需要的模型最小参数. 中发帖
用的是ollama+gemma，测试下来12b以上表达算比较流畅（量化后要求显存10G左右），1b和4b对于翻译的准确性来说，还是有点不够。
小说原文：
“Or else it’s the house. The house is haunted—or bewitched or something … I see things that have happened there—or else I see things that are going to happen there—and that would be worse. Perhaps a woman called Helen isgoing to be murdered there … Only I don’t see if it’s the house that’s haunted why I should see these...
────────── 链接信息 ──────────
🔗 论坛链接: linux.do
📎 访问地址: https://linux.do/t/topic/583233/1
─────────────────────────────
📢 来源：LINUX DO
