update

嘉渊

2023-04-24 6427c834dfd97b1f05c6659cdc7ccf010bf82fe1

 funasr/export/models/e2e_vad.py

@@ -11,7 +11,7 @@
class E2EVadModel(nn.Module):
    def __init__(self, model,
                max_seq_len=512,
                feats_dim=560,
                feats_dim=400,
                model_name='model',
                **kwargs,):
        super(E2EVadModel, self).__init__()
@@ -24,19 +24,10 @@
            raise "unsupported encoder"
        

    def forward(self, feats: torch.Tensor,
                       in_cache0: torch.Tensor,
                       in_cache1: torch.Tensor,
                       in_cache2: torch.Tensor,
                       in_cache3: torch.Tensor,
                       ):
    def forward(self, feats: torch.Tensor, *args, ):

        scores, cache0, cache1, cache2, cache3 = self.encoder(feats,
                                                              in_cache0,
                                                              in_cache1,
                                                              in_cache2,
                                                              in_cache3)  # return B * T * D
        return scores, cache0, cache1, cache2, cache3
        scores, out_caches = self.encoder(feats, *args)
        return scores, out_caches

    def get_dummy_inputs(self, frame=30):
        speech = torch.randn(1, frame, self.feats_dim)