{"provider_url":"https://hatena.blog","description":"ollama\u516c\u5f0f\u30e2\u30c7\u30eb\u306bgemma3:27b-it-qat\u304c\u52a0\u308f\u308a\u307e\u3057\u305f\u3002gemma3\u306f\u3001\u65e5\u672c\u8a9e\u3082\u5f37\u3044\u306e\u3067\u65e9\u901f\u8a66\u3057\u307e\u3059\u3002Colab L4\u74b0\u5883\u3067\u8a66\u3057\u307e\u3059\u3002VRAM\u306e\u5360\u6709\u7387\u306f18507MiB / 23034MiB\u3067\u3057\u305f\u3002\u3053\u306e\u30e2\u30c7\u30eb\u306f\u300cQuantization aware trained models (QAT)The quantization aware trained Gemma 3 models preserves similar quality as half precision models (BF16) while maintaining a lower memory footprint\u2026","version":"1.0","width":"100%","blog_url":"https://bwgift.hatenadiary.jp/","author_url":"https://blog.hatena.ne.jp/bwgift/","image_url":null,"height":"190","type":"rich","url":"https://bwgift.hatenadiary.jp/entry/2025/04/19/212236","author_name":"bwgift","title":"gemma3:27b-it-qat\u3092ollama\u3068colab\u3067\u8a66\u3057\u3066\u307f\u308b\u3002","categories":["LLM","Colaboratory","Python","\u81ea\u7136\u8a00\u8a9e\u51e6\u7406"],"blog_title":"\u5730\u5e73\u7dda\u307e\u3067\u884c\u3063\u3066\u304f\u308b\u3002","published":"2025-04-19 21:22:36","html":"<iframe src=\"https://hatenablog-parts.com/embed?url=https%3A%2F%2Fbwgift.hatenadiary.jp%2Fentry%2F2025%2F04%2F19%2F212236\" title=\"gemma3:27b-it-qat\u3092ollama\u3068colab\u3067\u8a66\u3057\u3066\u307f\u308b\u3002 - \u5730\u5e73\u7dda\u307e\u3067\u884c\u3063\u3066\u304f\u308b\u3002\" class=\"embed-card embed-blogcard\" scrolling=\"no\" frameborder=\"0\" style=\"display: block; width: 100%; height: 190px; max-width: 500px; margin: 10px 0px;\"></iframe>","provider_name":"Hatena Blog"}