I got a error

[wildboy2arthur]

Thanks for sharing, it's very interesting, I also want to make a .npy file. Follow your instructions to perform the installation step by step, making no mistakes until the last step.
My error message is as follows:

0%| | 0/1 [00:06<?, ?it/s]
Traceback (most recent call last):
File "D:\AniTalker\talking_face_preprocessing_back\extract_audio_features.py", line 52, in
main(args)
File "D:\AniTalker\talking_face_preprocessing_back\extract_audio_features.py", line 34, in main
outputs = model(input_values, output_hidden_states=True)
File "C:\Users\Fadawan.conda\envs\tfpw\lib\site-packages\torch\nn\modules\module.py", line 1110, in _call_impl
return forward_call(*input, **kwargs)
File "C:\Users\Fadawan.conda\envs\tfpw\lib\site-packages\transformers\models\hubert\modeling_hubert.py", line 1074, in forward
encoder_outputs = self.encoder(
File "C:\Users\Fadawan.conda\envs\tfpw\lib\site-packages\torch\nn\modules\module.py", line 1110, in _call_impl
return forward_call(*input, **kwargs)
File "C:\Users\Fadawan.conda\envs\tfpw\lib\site-packages\transformers\models\hubert\modeling_hubert.py", line 800, in forward
layer_outputs = layer(
File "C:\Users\Fadawan.conda\envs\tfpw\lib\site-packages\torch\nn\modules\module.py", line 1110, in _call_impl
return forward_call(*input, **kwargs)
File "C:\Users\Fadawan.conda\envs\tfpw\lib\site-packages\transformers\models\hubert\modeling_hubert.py", line 630, in forward
hidden_states, attn_weights, _ = self.attention(
File "C:\Users\Fadawan.conda\envs\tfpw\lib\site-packages\torch\nn\modules\module.py", line 1110, in _call_impl
return forward_call(*input, **kwargs)
File "C:\Users\Fadawan.conda\envs\tfpw\lib\site-packages\transformers\models\hubert\modeling_hubert.py", line 504, in forward
attn_weights = nn.functional.softmax(attn_weights, dim=-1)
File "C:\Users\Fadawan.conda\envs\tfpw\lib\site-packages\torch\nn\functional.py", line 1818, in softmax
ret = input.softmax(dim)
RuntimeError: CUDA out of memory. Tried to allocate 8.12 GiB (GPU 0; 12.00 GiB total capacity; 9.68 GiB already allocated; 0 bytes free; 10.74 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

[nitinmukesh]

Please wait for some time.
There is a pull request once merged the manual step of creating audio feature will not be required

#31

[wildboy2arthur]

Thank you! I am looking forward to.

[liutaocode]

Hello, it seems the GPU memory is full. May I ask what is the duration of your audio file? Additionally, to facilitate quick testing, you can first test with audio under 10 seconds or modify the Hubert audio extraction process to use CPU (slower but won't cause OOM). After extraction is complete, subsequent calls to Anitalker can normally use CUDA. (PS: The main GPU tested for this library is the 3090 with 24GB, which is larger than your 12GB, so some cases may not pass testing on your GPU.)

[nitinmukesh]

Manual process is removed and new WebUI
Please follow
https://github.com/X-LANCE/AniTalker/blob/main/md_docs/run_on_windows.md

[nitinmukesh]

Solution provided and no updates on this topic. Please open new issue if you face any issues.