espnet2.enh.layers.fasnet.test_model

source

espnet2.enh.layers.fasnet.test_model(model)

Test the given model with random input data.

This function generates random input data to test the functionality of the provided model. It creates a batch of audio signals with a specified number of microphones and verifies the output shape of the model for both variable and fixed microphone configurations.

• Parameters:model (nn.Module) – The model to be tested, which should implement a forward method that accepts input data and the number of active microphones.

Examples

>>> model_TAC = FaSNet_TAC(
...     enc_dim=64,
...     feature_dim=64,
...     hidden_dim=128,
...     layer=4,
...     segment_size=50,
...     nspk=2,
...     win_len=4,
...     context_len=16,
...     sr=16000,
... )
>>> test_model(model_TAC)
torch.Size([2, 2, 32000]) torch.Size([2, 2, 32000])

NOTE

The input tensor x is generated with a shape of (batch, num_mic, length), where batch is set to 2, num_mic is set to 4, and length is set to 32000. The num_mic tensor is created as an ad-hoc array representing the number of active microphones for each batch element.

• Raises:AssertionError – If the model’s output shape does not match the expected dimensions.