Traceback (most recent call last):
File "C:\Users\averw\Desktop\repos\vocal-remover\inference.py", line 184, in <module>
main()
File "C:\Users\averw\Desktop\repos\vocal-remover\inference.py", line 152, in main y_spec, v_spec = sp.separate(X_spec)
File "C:\Users\averw\Desktop\repos\vocal-remover\inference.py", line 77, in separate
mask = self._separate(X_mag_pad, roi_size)
File "C:\Users\averw\Desktop\repos\vocal-remover\inference.py", line 42, in _separate
X_batch = torch.from_numpy(X_batch).to(self.device)
RuntimeError: Numpy is not available
why is this error happening in PyTorch? I have numpy 1.21.2 installed, and I have PyTorch Nightly with CUDA 11.3
python -c "import torch; import numpy as np; print(torch.from_numpy(np.random.randn(1)))"
Traceback (most recent call last):
File "<string>", line 1, in <module>
ModuleNotFoundError: No module named 'numpy'
I don’t think the linked issue is Linux-specific and would assume it’s also tracking Windows support.
In any case, are you seeing any issues using the Python 3.10 binaries?
are you seeing any issues using the Python 3.10 binaries?
@ptrblck Yes (if I’m not missing something). We’re trying to enable testing with Python 3.10 on Linux/macOS/Windows in CI, but it fails only on Windows due to the same error as reported here.
@ptrblck I’m confused about whether it’s supported by PyTorch or not, because, as @averwhy mentioned above, there is a statement about Python 3.10 on Windows in the official instruction page: Start Locally | PyTorch
Currently, PyTorch on Windows only supports Python 3.7-3.9; Python 2.x is not supported.
I don’t fully understand the linked issue by skimming through it.
It seems you depend on numpy but fail to build it.
However, in your environment you already have numpy installed so could you explain how the numpy issue is related to PyTorch’s Python 3.10 support?