新版本模型训练问题与解决方法

新版本主目录里没有synthesizer_train.py，查找了一下发现...\control\cli里面有synthesizer_train.py文件，但是运行称缺一个models的依赖，pip安装models提示需要base依赖，遂安装base，提示gbk编码器错误，按照网上教程为base的setup.py第21行指定编码器为utf-8依旧报错，遂放弃使用命令行进行训练，改用web中自带的模型训练功能进行训练，正常，然后关闭，再启动命令行训练，即可正常使用。

Start by comparing the missing root-level synthesizer_train.py with control/cli/synthesizer_train.py and reproduce the reported command-line training setup. Inspect the dependency installation failure involving models and base, including line 21 of base/setup.py and the GBK encoding error. Done means command-line training works without the reported dependency or encoding errors, matching the web training path.

Written by the indexing model from the issue text.