[GH-ISSUE #414] MacOS Apple M1无法训练won't complete the the training process #211

New issue

Closed

opened 2026-10-20 09:56:45 +00:00 by bsakaguchi · 5 comments

bsakaguchi commented

2026-10-20 09:56:45 +00:00

Owner

Originally created by @eisneim on GitHub (Jun 2, 2023).
Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/414

所有依赖都装好了，前两部：预处理，提取特征，得到3_feature256文件夹都没有问题，但是到了真正训练的时候就一瞬间完成了，直接跳过了，batch_size for every GPU: 1 各种参数都试过都不行，电机训练模型，一点击基本上就显示完成

step 1: processing data

...这里省略

step 3a: model traning started
write filelist done
python train_nsf_sim_cache_sid_load_pretrain.py -e eisneim -sr 40k -f0 1 -bs 1 -g 0 -te 100 -se 5 -pg pretrained/f0G40k.pth -pd pretrained/f0D40k.pth -l 0 -c 0 -sw 1 -v v1
Training complete. Logs are available in the console, or the 'train.log' under experiment folder
(50590, 256),1297
training index
adding index
成功构建索引, added_IVF1297_Flat_nprobe_1_eisneim_v1.index
all processes have been completed!

log文件里没有ckpt或者pth文件
是不是在Mac Arm平台下完全不支持训练？

Originally created by @eisneim on GitHub (Jun 2, 2023). Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/414 所有依赖都装好了，前两部：预处理，提取特征，得到`3_feature256`文件夹都没有问题，但是到了真正训练的时候就一瞬间完成了，直接跳过了，*batch_size for every GPU: 1* 各种参数都试过都不行，电机训练模型，一点击基本上就显示完成 ``` step 1: processing data ...这里省略 step 3a: model traning started write filelist done python train_nsf_sim_cache_sid_load_pretrain.py -e eisneim -sr 40k -f0 1 -bs 1 -g 0 -te 100 -se 5 -pg pretrained/f0G40k.pth -pd pretrained/f0D40k.pth -l 0 -c 0 -sw 1 -v v1 Training complete. Logs are available in the console, or the 'train.log' under experiment folder (50590, 256),1297 training index adding index 成功构建索引, added_IVF1297_Flat_nprobe_1_eisneim_v1.index all processes have been completed! ``` log文件里没有ckpt或者pth文件是不是在Mac Arm平台下完全不支持训练？

bsakaguchi

2026-10-20 09:56:45 +00:00

closed this issue
added the
enhancement

duplicate
labels

bsakaguchi commented

2026-10-20 09:56:46 +00:00

Author

Owner

@Tps-F commented on GitHub (Jun 2, 2023):

now i'm progressing
#313

@Tps-F commented on GitHub (Jun 2, 2023): now i'm progressing #313

bsakaguchi commented

2026-10-20 09:56:46 +00:00

Author

Owner

@eisneim commented on GitHub (Jun 3, 2023):

here is my quick solution and it works: change just one line in train_nsf_sim_cache_sid_load_pretrain.py

def main():
     n_gpus = torch.cuda.device_count()

change to this:

def main():
     n_gpus = torch.backends.mps.is_available() ? 1 : torch.cuda.device_count()

and it works, here are some ouputs

...
INFO:eisneim:loss_disc=2.876, loss_gen=2.512, loss_fm=8.741,loss_mel=18.739, loss_kl=1.106
INFO:eisneim:====> Epoch: 67 [2023-06-03 06:55:52] | (0:10:21.757121)
INFO:eisneim:====> Epoch: 68 [2023-06-03 07:06:17] | (0:10:24.585755)
INFO:eisneim:====> Epoch: 69 [2023-06-03 07:16:43] | (0:10:25.606220)
INFO:eisneim:Train Epoch: 70 [84%]

@eisneim commented on GitHub (Jun 3, 2023): here is my quick solution and it works: change just one line in `train_nsf_sim_cache_sid_load_pretrain.py` ``` def main(): n_gpus = torch.cuda.device_count() ``` change to this: ``` def main(): n_gpus = torch.backends.mps.is_available() ? 1 : torch.cuda.device_count() ``` and it works, here are some ouputs ``` ... INFO:eisneim:loss_disc=2.876, loss_gen=2.512, loss_fm=8.741,loss_mel=18.739, loss_kl=1.106 INFO:eisneim:====> Epoch: 67 [2023-06-03 06:55:52] | (0:10:21.757121) INFO:eisneim:====> Epoch: 68 [2023-06-03 07:06:17] | (0:10:24.585755) INFO:eisneim:====> Epoch: 69 [2023-06-03 07:16:43] | (0:10:25.606220) INFO:eisneim:Train Epoch: 70 [84%] ```

bsakaguchi commented

2026-10-20 09:56:46 +00:00

Author

Owner

@Tps-F commented on GitHub (Jun 3, 2023):

Sure, i know
but It will be a temporary fix, That may cause a permanent loop to continue in some environments.
I'm trying to figure out what caused that...

@Tps-F commented on GitHub (Jun 3, 2023): Sure, i know but It will be a temporary fix, That may cause a permanent loop to continue in some environments. I'm trying to figure out what caused that...

bsakaguchi commented

2026-10-20 09:56:46 +00:00

Author

Owner

@Tps-F commented on GitHub (Jun 3, 2023):

I think I might have figured out why.

@Tps-F commented on GitHub (Jun 3, 2023): I think I might have figured out why.

bsakaguchi commented

2026-10-20 09:56:46 +00:00

Author

Owner

@Tps-F commented on GitHub (Jun 5, 2023):