v5.8.0

更新日志

• 更新GalTransl-7B-v2.5模型，抑制了一些已知问题，并且在对比v2时文风更细腻

• 重构部分代码，现在支持边翻译边输出已完成的文件

• GPT字典现在会自动去重

• sakura: 增加设置项gpt.token_limit：大于0时限制每轮请求的token数量。防止爆context时可以设置为llama.cpp -c的数值除2。

• sakura: 现在设置目标语言为繁体中文时会自动进行转换

• sakura: 修正退化检测的bug，导致在某些情况下反复翻译

• 修正缓存的一处bug，在特定情况下导致反复翻译

• 翻译非galgame文件时，不去隐藏每行的日文对话框

• 优化示例配置文件

• Epub文件插件：修复开启分割后，纯图片单文件无法正常输出导致的epub合并失败的问题 by @PiDanShouRouZhouXD

• translator++文件插件： 修复ILLEGAL_CHARACTERS问题 by @PiDanShouRouZhouXD

Changelog

• Updated GalTransl-7B-v2.5 model, suppressing some known issues and achieving a more nuanced writing style compared to v2

• Refactored parts of the code, now supporting output of completed files while translation is in progress

• GPT dictionary now automatically removes duplicates

• sakura: Added setting gpt.token_limit: When greater than 0, it limits the number of tokens per request round. To prevent context overflow, it can be set to half the value of llama.cpp -c

• sakura: Now automatically converts to Traditional Chinese when the target language is set to Traditional Chinese

• sakura: Fixed a bug in degradation detection that caused repeated translations in certain situations

• Fixed a caching bug that caused repeated translations in specific cases

• When translating non-galgame files, Japanese dialogue boxes for each line are no longer hidden

• Optimized the sample configuration file

• Epub file plugin: Fixed an issue where pure image single files couldn't be output normally when splitting was enabled, causing epub merging to fail (by @PiDanShouRouZhouXD)

• translator++ file plugin: Fixed ILLEGAL_CHARACTERS problem (by @PiDanShouRouZhouXD)

---

GalTransl-ver：脚本完整包/Script complete package
GalTransl-ver-core：脚本更新包，不会覆盖你的字典/Script update package, will not overwrite your dictionary
GalTransl-ver-win：Windows免环境完整包/Windows environment-free complete package