模型介绍

Sakura-GalTransl-14B-v3基于Sakura-14B-Qwen2.5-v1.0,是GalTransl-v3模型的14b版本

得益于更大的底模及改进的对齐训练,GalTransl-14B-v3整体质量好于GalTransl-7B-v3

详细介绍及使用方式等可以参考Sakura-GalTransl-7B-v3

推荐温度0.3,top_p 0.8


autodl部署教程

参考https://books.fishhawk.top/forum/65719bf16843e12bd3a4dc98 ,选择4090部署FishHawk/auto-novel/Sakura-LNovel v1.0.0版镜像

打开JupyterLab后,按顺序输入以下2个命令

  1. 下载模型:
wget https://hf-mirror.com/SakuraLLM/Sakura-GalTransl-14B-v3/resolve/main/Sakura-Galtransl-14B-v3-Q5_K_S.gguf   
  1. 不要按教程运行.run,而是运行下面的命令:
./llama.cpp/llama-server -m Sakura-Galtransl-14B-v3-Q5_K_S.gguf -c 32768 -np 16 -fa -ngl 999 -a Sakura-Galtransl-14B-v3-Q5_K_S --port 6006   

然后继续按"SSH转发服务到本地"操作即可。接口地址为http://127.0.0.1:6006

Downloads last month
3,589
GGUF
Model size
14.8B params
Architecture
qwen2
Hardware compatibility
Log In to view the estimation

4-bit

5-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support