Plachtaa / seed-vc

State-of-the-Art zero-shot voice conversion & singing voice conversion with in context learning
GNU General Public License v3.0
662 stars 76 forks source link

推理速度超级慢?! #48

Open chaorenai opened 1 week ago

chaorenai commented 1 week ago

我用的是4090,但是GPU只用了一点点,效果很棒,但是推理速度太慢了。是之前使用rvc模型推理时间的100倍都不止……是我哪里用错了吗?

Plachtaa commented 1 week ago

如果你需要debug帮助,希望你可以提供:

  1. 使用的是哪个script
  2. 是否是singing voice conversion
  3. torch.cuda.available()是否返回True
  4. source & target音频文件
teressawang commented 1 week ago

我也是,推理速度才 2.58it/s , 用的 是app.py 的推理脚本, T4 ,torch.cuda.available() 返回True

teressawang commented 1 week ago

Guessed Channel Layout for Input Stream #0.0 : stereo Input #0, wav, from 'donnachen.wav': Metadata: encoder : Lavf58.45.100 Duration: 00:00:20.15, bitrate: 1411 kb/s Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 44100 Hz, stereo, s16, 1411 kb/s

Plachtaa commented 1 week ago

我也是,推理速度才 2.58it/s , 用的 是app.py 的推理脚本, T4 ,torch.cuda.available() 返回True

T4上这个速度是正常的,svc的模型本身参数也更多

Nuyoah111111 commented 1 week ago

我也是,推理速度才 2.58it/s , 用的 是app.py 的推理脚本, T4 ,torch.cuda.available() 返回True

T4上这个速度是正常的,svc的模型本身参数也更多

我也是运行的app.py这个脚本 我是v100的gpu 速度也很慢

zhixianjuli commented 2 days ago

You can uninstall torch torchvision torchaudio. And then install as
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

teressawang commented 2 days ago

You can uninstall torch torchvision torchaudio. And then install as pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

not work for me ,but thx~