RVC-Project/Retrieval-based-Voice-Conversion-WebUI

Fork 0

mirror of https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI.git synced 2026-11-05 00:27:30 +00:00

[GH-ISSUE #415] onnx_inference_demo how do I get the vec-256-layer-9? #212

New issue

Closed

opened 2026-10-20 09:56:45 +00:00 by bsakaguchi · 7 comments

bsakaguchi commented

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

Owner

Originally created by @mathigatti on GitHub (Jun 2, 2023).
Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/415

Hi! I'm trying to test the onnx_inference_demo.py script, I created a model.onnx using the export_onnx.py script and set that as model_path. But now it looks like I'm missing that "vec_name" thing, not sure how should I generate that file.

These are lines I'm talking about at onnx_inference_demo.py

model_path = "model.onnx"  # 模型的完整路径
vec_name = "vec-256-layer-9"  # 内部自动补齐为 f"pretrained/{vec_name}.onnx" 需要onnx的vec模型

Originally created by @mathigatti on GitHub (Jun 2, 2023). Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/415 Hi! I'm trying to test the [onnx_inference_demo.py](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/onnx_inference_demo.py) script, I created a model.onnx using the export_onnx.py script and set that as model_path. But now it looks like I'm missing that "vec_name" thing, not sure how should I generate that file. These are lines I'm talking about at onnx_inference_demo.py ```python model_path = "model.onnx" # 模型的完整路径 vec_name = "vec-256-layer-9" # 内部自动补齐为 f"pretrained/{vec_name}.onnx" 需要onnx的vec模型 ```

bsakaguchi closed this issue

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

bsakaguchi commented

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

Author

Owner

@mathigatti commented on GitHub (Jun 2, 2023):

okay it looks like it's contentVec ported to onnx, I just found it here.

@mathigatti commented on GitHub (Jun 2, 2023): okay it looks like it's contentVec ported to onnx, I just found it [here](https://huggingface.co/NaruseMioShirakana/MoeSS-SUBModel/tree/main).

bsakaguchi commented

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

Author

Owner

@SirHumphreyforreal commented on GitHub (Mar 1, 2024):

sorry to interrupt,I encounter the same thing as you did,I follow your link to download this"vec-256-layer-9" thing ,but it keeps telling me
"Load model from pretrained/vec-256-layer-9.onnx failed:Load model pretrained/vec-256-layer-9.onnx failed. File doesn't exist"
wherever I store it, even I use absolute path.
I really dont know what goes wrong,pls help me

@SirHumphreyforreal commented on GitHub (Mar 1, 2024): sorry to interrupt,I encounter the same thing as you did,I follow your link to download this"vec-256-layer-9" thing ,but it keeps telling me "Load model from pretrained/vec-256-layer-9.onnx failed:Load model pretrained/vec-256-layer-9.onnx failed. File doesn't exist" wherever I store it, even I use absolute path. I really dont know what goes wrong,pls help me

bsakaguchi commented

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

Author

Owner

@mathigatti commented on GitHub (Mar 1, 2024):

share your code

@mathigatti commented on GitHub (Mar 1, 2024): share your code

bsakaguchi commented

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

Author

Owner

@SirHumphreyforreal commented on GitHub (Mar 2, 2024):

Here is the code:

import soundfile
from ..infer.lib.infer_pack.onnx_inference import OnnxRVC

hop_size = 512
sampling_rate = 40000 # 采样率
f0_up_key = 0 # 升降调
sid = 0 # 角色ID
f0_method = "dio" # F0提取算法
model_path = "test.onnx" # 模型的完整路径
vec_name = (
"vec-256-layer-9.onnx" # 内部自动补齐为 f"pretrained/{vec_name}.onnx" 需要onnx的vec模型
)
wav_path = "sunflower.wav" # 输入路径或ByteIO实例
out_path = "out.wav" # 输出路径或ByteIO实例

model = OnnxRVC(
model_path, vec_path=vec_name, sr=sampling_rate, hop_size=hop_size, device="cuda"
)

audio = model.inference(wav_path, sid, f0_method=f0_method, f0_up_key=f0_up_key)

soundfile.write(out_path, audio, sampling_rate)

and output in terminal as follows:

Traceback (most recent call last):
File "D:\QT\test.py", line 17, in
model = OnnxRVC(
File "D:\QT\onnx.py", line 74, in init
self.vec_model = ContentVec(vec_path, device)
File "D:\QT\onnx.py", line 22, in init
self.model = onnxruntime.InferenceSession(vec_path, providers=providers)
File "C:\Python38\lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 419, in init
self._create_inference_session(providers, provider_options, disabled_optimizers)
File "C:\Python38\lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 472, in _create_inference_session
sess = C.InferenceSession(session_options, self._model_path, True, self._read_config_from_model)
onnxruntime.capi.onnxruntime_pybind11_state.NoSuchFile: [ONNXRuntimeError] : 3 : NO_SUCHFILE : Load model from pretrained/vec-256-layer-9.onnx failed:Load model pretrained/vec-256-layer-9.onnx failed. File doesn't exist

@SirHumphreyforreal commented on GitHub (Mar 2, 2024): Here is the code: import soundfile from ..infer.lib.infer_pack.onnx_inference import OnnxRVC hop_size = 512 sampling_rate = 40000 # 采样率 f0_up_key = 0 # 升降调 sid = 0 # 角色ID f0_method = "dio" # F0提取算法 model_path = "test.onnx" # 模型的完整路径 vec_name = ( "vec-256-layer-9.onnx" # 内部自动补齐为 f"pretrained/{vec_name}.onnx" 需要onnx的vec模型 ) wav_path = "sunflower.wav" # 输入路径或ByteIO实例 out_path = "out.wav" # 输出路径或ByteIO实例 model = OnnxRVC( model_path, vec_path=vec_name, sr=sampling_rate, hop_size=hop_size, device="cuda" ) audio = model.inference(wav_path, sid, f0_method=f0_method, f0_up_key=f0_up_key) soundfile.write(out_path, audio, sampling_rate) and output in terminal as follows: Traceback (most recent call last): File "D:\QT\test.py", line 17, in <module> model = OnnxRVC( File "D:\QT\onnx.py", line 74, in __init__ self.vec_model = ContentVec(vec_path, device) File "D:\QT\onnx.py", line 22, in __init__ self.model = onnxruntime.InferenceSession(vec_path, providers=providers) File "C:\Python38\lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 419, in __init__ self._create_inference_session(providers, provider_options, disabled_optimizers) File "C:\Python38\lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 472, in _create_inference_session sess = C.InferenceSession(session_options, self._model_path, True, self._read_config_from_model) onnxruntime.capi.onnxruntime_pybind11_state.NoSuchFile: [ONNXRuntimeError] : 3 : NO_SUCHFILE : Load model from pretrained/vec-256-layer-9.onnx failed:Load model pretrained/vec-256-layer-9.onnx failed. File doesn't exist

bsakaguchi commented

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

Author

Owner

@mathigatti commented on GitHub (Mar 2, 2024):

How are you running it and where is the pretrained folder located? I think you should run it from the source folder (python tools/onnx_inference_demo.py) of the project and place the pretrained folder there in the root directory too

@mathigatti commented on GitHub (Mar 2, 2024): How are you running it and where is the pretrained folder located? I think you should run it from the source folder (python tools/onnx_inference_demo.py) of the project and place the pretrained folder there in the root directory too

bsakaguchi commented

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

Author

Owner

@ZzAR0 commented on GitHub (Mar 3, 2024):

Hi, I have a similar problem, I have vec-256-layer-9.onnx under the same file as vec-768-layer-12.onnx.
But there was an error：onnxruntime.capi.onnxruntime_pybind11_state.InvalidArgument: [ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Got invalid dimensions for input: phone for the following indices
index: 2 Got: 256 Expected: 768 Please fix either the inputs/outputs or the model.

@ZzAR0 commented on GitHub (Mar 3, 2024): Hi, I have a similar problem, I have vec-256-layer-9.onnx under the same file as vec-768-layer-12.onnx. But there was an error：onnxruntime.capi.onnxruntime_pybind11_state.InvalidArgument: [ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Got invalid dimensions for input: phone for the following indices index: 2 Got: 256 Expected: 768 Please fix either the inputs/outputs or the model.

bsakaguchi commented

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

Author

Owner

@ZzAR0 commented on GitHub (Mar 3, 2024):

Also ask why the onnx model for inference works so poorly and whether onnx can now use the rmvpe algorithm Thanks.

@ZzAR0 commented on GitHub (Mar 3, 2024): Also ask why the onnx model for inference works so poorly and whether onnx can now use the rmvpe algorithm Thanks.