本模组通过采用自适应马尔可夫链算法来学习通信模式,从而在 Minecraft 聊天中实现更快的打字速度和更少的拼写错误。自动补全功能起初可能显得有些笨拙,但随着数据积累,其表现会逐渐提升。
语言:
默认仅支持英语。若要更改模组的语言,请前往 ...\AppData\Roaming.minecraft\config\dictionary.json 。找到 dictionary.json 文件后,你可以将其替换为你所需的词典,只要该文件是符合 JSON 格式且包含 words:integer(默认为 0) 的 JSON 文件即可。
LLM 模式:
模组内置一个 LLM 引擎,默认处于禁用状态。启用后,它需要下载相关依赖项,稍等片刻即可就绪。如果下载完成后建议仍无法正常工作,请关闭游戏,进入 ...\AppData\Roaming.minecraft\mods 目录并删除名为 llm 的文件夹,然后重新启动游戏。我不推荐使用 LLM 模式,因为它速度较慢,且效果并不比经过适度训练的马尔可夫链更好。
模组配置:
默认情况下,打开模组配置界面的快捷键为“k”;你可以在 Minecraft 基础设置的“控制”选项卡中更改此键位。
按下该键后,将打开配置界面:
- Engine(引擎) 可设置为 Markov(建议使用此项)或 LLM。选择 LLM 模式时,需在 CPU 和 CUDA 处理之间做出选择。CUDA 仅适用于 NVIDIA GPU,但速度远快于 CPU。
- Reset Trie weights(重置前缀树权重) 会将你的词典重置为未训练过的英文状态。
- Reset Markov chain(重置马尔可夫链) 将清除马尔可夫链的训练完整性。
- Max suggestions(最大建议数) 范围为 0 到 99,表示 Markov 模式下建议列表的最大数量。在 LLM 模式下,最大建议数代表 LLM 输出的最大单词长度。
- Minimum word length(最小单词长度) 是你希望马尔可夫链和词典输出的最小单词长度。
- Chat Autocomplete(聊天自动补全) 用于启用或禁用该模组。
- Save(保存) 将重启引擎,因此按下时会出现短暂卡顿。若要不保存退出,请按 escape。
重要提示: 马尔可夫链功能可能会通过聊天保存玩家的完整消息。这些本地保存的数据仅用于提升用户体验,绝不会与他人共享。保存的内容可在 Minecraft 配置文件中找到:26.2 版本位于 autocomplete_markov.json,26.3 版本位于 markov_chain.json。请注意,如果长时间使用该模组,文件大小可能会超过 1 MB。
This mod allows faster typing and fewer typos in the minecraft chat with an adaptive Markov chain algorithm that learns communication patterns. The autocomplete might feel stupid at first, but as it gets more data, it should get better.
Language:
The only language supported by default is English. To change the mod's language, go to ...\AppData\Roaming.minecraft\config\dictionary.json Once you have located the dictionary.json file, you can replace it with your wanted dictionary as long as it is a json file containing words:integer(0 by default) in the json format.
LLM mode:
The mod contains an LLM engine that is disabled by default. Once you turn it on. It should download the requirements and be ready after a bit. If the suggestions don’t work even after the download is completed, close your game, go to ...\AppData\Roaming.minecraft\mods and delete the folder called llm. Then reboot your game. I do not recommend using LLM mode, as it is slow and not really better than a medium trained Markov chain.
Mod config:
By default, the key bind to open the mod's config screen is “k”; it can be changed in the controls tab of the base minecraft settings.
When the key is pressed, a config screen will open:
- Engine can be set to Markov (you should take that one) or LLM. When LLM mode is selected, you have to choose between CPU or CUDA processing. CUDA only works for NVIDIA GPUs but is way faster than CPU.
- Reset Trie weights will reset your dictionary to untrained English.
- Reset Markov chain will wipe the integrity of the Markov chain's training.
- Max suggestions are from 0 to 99 and is the max size of the suggested list in Markov mode. In LLM mode, max suggestions represent the max length in words the LLM will output.
- Minimum word length is the smallest word size you want the Markov chain and dictionary to output.
- Chat Autocomplete is to enable/disable the mod.
- Save will restart the engine, so a small freeze will occur when pressed. To exit without saving, press on escape.
IMPORTANT: The Markov chain function could save entire messages from players through the chat. That locally saved data is used ONLY for a better user experience and will NEVER be shared. The saved contents can be found in the minecraft config files at autocomplete_markov.json for the 26.2 version and at markov_chain.json for the 26.3 version. Note that the file could end up bigger than a MB of storage if the mod is used for a long time.
评论 0
登录后参与评分
还没有评论,来抢沙发吧