search

fakufaku / fast_bss_eval

A fast implementation of bss_eval metrics for blind source separation
https://fast-bss-eval.readthedocs.io/en/latest/
MIT License
130 stars 8 forks source link

RuntimeError: min(): Expected reduction dim to be specified for input.numel() == 0 #17

Closed dr-costas closed 1 year ago

dr-costas commented 1 year ago

Hi,

I'm trying to use the torch.bss_eval_sources and I'm getting the error:

fast_bss_eval/torch/helpers.py", line 142, in _linear_sum_assignment_with_inf 
  m = values.min()
RuntimeError: min(): Expected reduction dim to be specified for input.numel() == 0. Specify the reduction dim with the 'dim' argument.

Any pointers on that?

fakufaku commented 1 year ago

Hi @dr-costas Did you figure it out ?

dr-costas commented 1 year ago

Hi,

Well, no, but I side-stepped it. The problem, n my case, occurs when trying to calculate permutations, which I really don't need. So, I don't calculate permutations and all are good :)

fakufaku commented 1 year ago

Ok 😄 I think this may happen when one of the references is all zeros, or if you have some nan or inf in your signal.