Downloads · 30 days
10.9K
11% of all-time downloads
SakuraLLM/Sakura-GalTransl-14B-v3.8
Sakura-GalTransl-14B-v3.8 is a machine learning model from SakuraLLM. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as cc-by-nc-sa-4.0.
基于Sakura-14B-Qwen2.5-v1.0,是GalTransl-v3模型的14b版本
Downloads · 30 days
10.9K
11% of all-time downloads
All-time downloads
100K
Public
Repo size
76.4 GB
Likes
42
Public
Click a slice to open those files.
.gguf76.4 GB · 100%
From the Hugging Face model README
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,GalTransl-14B整体质量好于GalTransl-7B
本地部署方式和prompt模板等可以参考Sakura-GalTransl-7B-v3
推荐温度0.3,top_p 0.8
2025.9 v3.8:改进文风,修正了部分已知的控制符号保留问题。注:不带后缀的版本量化等级为Q6_K
2025.7 v3.7:通过改进的强化学习,显著改进了整体质量并减少人称错误
2025.5 v3:初版
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
wget https://hf-mirror.com/SakuraLLM/Sakura-GalTransl-14B-v3.8/resolve/main/Sakura-Galtransl-14B-v3.8.gguf
./llama.cpp/llama-server -m Sakura-Galtransl-14B-v3.8.gguf -c 32768 -np 16 -fa -ngl 999 -a Sakura-Galtransl-14B-v3.8 --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 ,翻译时可以设置16线程最大化速度,16线程跑满时每小时大约翻译3w句