Bella

请问底膜hubert_base.pt转成hubert_base.onnx,你们实现了吗？

> Very very great discussion! > > I converted huggingface model to onnx. Size is very small (280MB) ! And it work fine in my app that is [realtime voice...

Where can I get the v2 version of hubert_base.onnx [256] ?

C:\Users\Administrator\Desktop\OpenVoice\OpenVoice-main>python C:\Users\Administrator\Desktop\OpenVoice\OpenVoice-main\role_tts.py Loaded checkpoint 'checkpoints_v2/converter/checkpoint.pth' missing/unexpected keys: [] [] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssss OpenVoice version: v2 C:\AudioDspFolder\Zero-Shot\6600df81-cb77-4d60-85f9-34960c5d87bb-4785.mp3 C:\Users\Administrator\AppData\Local\Programs\Python\Python39\lib\site-packages\torch\functional.py:641: UserWarning: stft with return_complex=False is deprecated. In a future pytorch release, stft will return complex...

另外一个文本转语音的开源项目使用了融合cuda技术来实现推理加速，这个项目是否能够探索一下使用这个技术的可行性。 https://github.com/fishaudio/fish-speech/blob/main/docs/zh/inference.md 您可能希望使用 --compile 来融合 cuda 内核以实现更快的推理 (~30 个 token/秒 -> ~500 个 token/秒). 对应的, 如果你不打算使用加速, 你可以注释掉 --compile 参数. @Plachtaa

那个项目我也在Windows上尝试了一下，但是在compile的时候报了语法错误，不过我目前还是认为是我的cuda环境配置问题引起的。

def load_models(args): global fp16 fp16 = args.fp16 if not args.f0_condition: if args.checkpoint is None: dit_checkpoint_path, dit_config_path = load_custom_model_from_hf("Plachta/Seed-VC", "DiT_seed_v2_uvit_whisper_small_wavenet_bigvgan_pruned.pth", "config_dit_mel_seed_uvit_whisper_small_wavenet.yml") else: dit_checkpoint_path = args.checkpoint dit_config_path = args.config f0_fn = None...