Jonas Haag
Jonas Haag
> Please, do invest time if you feel like it Of course I didn’t get around to doing it 🙄 probably also not this week.
The problem is that ONNX doesn't support PyTorch's complex numbers. There isn't a lot we can do except for creating a facade for PyTorch's complex numbers that its based on...
Unfortunately I don't have access to the code for a few days. The approach I've taken is as follows: - Move all `t.abs()`, `t.angle()`, `t.view_as_complex()` etc calls on complex tensors...
Recommended by original author here https://github.com/ludlows/python-pesq/issues/12#issuecomment-645097746
No idea, sorry
Or simply switch to https://github.com/vBaiCai/python-pesq, which I have worked with in the past and didn’t cause any trouble
You're right, I just checked, the vBaiCai version has modified PESQ source. Maybe I'll create my own wrapper... I mean the wrapper source codes look really simple
What are we going to do with this problem?
I did a patch that adds the doc hint and the function. I realised the function may be generally useful, even to non-STFT models, to return the reconstructed size (without...
Looks good to me, but I wonder why roll our own? We could use something like https://github.com/bw2/ConfigArgParse instead (never used it but looks good). For my use cases, I think...