search

babysor / MockingBird

🚀AI拟声: 5秒内克隆您的声音并生成任意语音内容 Clone a voice in 5 seconds to generate arbitrary speech in real-time
Other
34.88k stars 5.18k forks source link

声音编码训练何时能结束？ #866

Open hujb2000 opened 1 year ago

hujb2000 commented 1 year ago

声音编码训练运行一下命令： C:\ProgramData\Anaconda3\envs\mockingbird\python.exe E:\workspace\MockingBird\control\cli\encoder_train.py my_run E:\datasets\SV2TTS\encoder -m E:\workspace\MockingBird\data\ckpt\encoder --visdom_server http://localhost ，当进行到step:44k,lose:0.1735,Err:0.0144,再收敛会不会过拟合? 微信图片_20230331145004 微信图片_20230331144947

babysor commented 1 year ago

还好，看图，跑到70k应该都没啥问题

hujb2000 commented 1 year ago

训练到这个程度: Step 75690 Loss: 0.0675 EER: 0.0063 Step time: mean: 1013ms std: 950ms 可以终止了吗,?手动强制停止程序没问题吧？

微信图片_20230331233718 微信图片_20230331233735

zhaozhao678 commented 1 year ago

您好，我是个新手，好多地方不太明白，我想问一下您如何训练编码器的，我想通过训练编码器来获得我的音频数据的特征向量，我该如何做呢？

HaSaKiYasuooo commented 1 year ago

@hujb2000 同样问题，如何训练编码器 (mocking2) F:\LSX\MockingBird>python F:\LSX\MockingBird\control\cli\encoder_train.py qh F:\LSX\MockingBird\SV2TTS\encoder -m F:\LSX\MockingBird\data\ckpt\encoder --no_visdom Arguments: run_id: qh clean_data_root: F:\LSX\MockingBird\SV2TTS\encoder models_dir: F:\LSX\MockingBird\data\ckpt\encoder vis_every: 10 umap_every: 100 save_every: 500 backup_every: 7500 force_restart: False visdom_server: http://localhost no_visdom: True

No model "qh" found, starting training from scratch.

HaSaKiYasuooo commented 1 year ago

您好，我是个新手，好多地方不太明白，我想问一下您如何训练编码器的，我想通过训练编码器来获得我的音频数据的特征向量，我该如何做呢？

你完成了如何训练编码器的命令吗，我试了很多次没有用