Use Pretrained Model for multi-channel input

[Raphael310]

Hello Asteroid-Team/Community,
I have a task where i need to separate 2-3 sound-sources from a noisy two channeled (binaural-audio) input. I am using pretrained asteroid models (like https://huggingface.co/JorisCos/ConvTasNet_Libri3Mix_sepnoisy_16k). But when i try to separate the 2 channels together i get an error. If i iterate over the channels and separate each separately it works but i want to handle them together to get better results.
I looked up into the class BaseModel where it looks like the in_channels are fixed to 1 because the in_channels argument isnt called from the init function from the inherited class. Can anyone explain me if it is possible to use mutli-channel input for pretrained models and how this is possible?

class Base(torch.nn.Module):
    """Base class for serializable models.

    Defines saving/loading procedures, and separation interface to `separate`.
    Need to overwrite the `forward` and `get_model_args` methods.

    Models inheriting from `BaseModel` can be used by :mod:`asteroid.separate`
    and by the `asteroid-infer` CLI. For models whose `forward` doesn't go from
    waveform to waveform tensors, overwrite `forward_wav` to return
    waveform tensors.

    Args:
        sample_rate (float): Operating sample rate of the model.
        in_channels: Number of input channels in the signal.
            If None, no checks will be performed.
    """

    def __init__(self, sample_rate: float, in_channels: Optional[int] = 1):
        super().__init__() 

[mpariente]

We do not have pretrained models for multichannel inputs. So splitting the inputs in channels is still the best way to do.