Merge branch 'dev_cmz2' of github.com:alibaba-damo-academy/FunASR into dev_...

游雁

2023-03-30 c5acc04e2df3316c284c3ab75575498934314560

 funasr/export/models/vad_realtime_transformer.py

@@ -21,7 +21,9 @@
        **kwargs,
    ):
        super().__init__()

        onnx = False
        if "onnx" in kwargs:
            onnx = kwargs["onnx"]

        self.embed = model.embed
        if isinstance(model.encoder, SANMVadEncoder):
@@ -30,6 +32,7 @@
            assert False, "Only support samn encode."
        # self.encoder = model.encoder
        self.decoder = model.decoder
        self.model_name = model_name



@@ -44,7 +47,7 @@
        """
        x = self.embed(input)
        # mask = self._target_mask(input)
        h, _, _ = self.encoder(x, text_lengths, vad_indexes)
        h, _ = self.encoder(x, text_lengths, vad_indexes)
        y = self.decoder(h)
        return y

@@ -53,12 +56,13 @@

    def get_dummy_inputs(self):
        length = 120
        text_indexes = torch.randint(0, self.embed.num_embeddings, (2, length))
        text_lengths = torch.tensor([length-20, length], dtype=torch.int32)
        return (text_indexes, text_lengths)
        text_indexes = torch.randint(0, self.embed.num_embeddings, (1, length))
        text_lengths = torch.tensor([length], dtype=torch.int32)
        vad_mask = torch.ones(length, length, dtype=torch.float32)[None, None, :, :]
        return (text_indexes, text_lengths, vad_mask)

    def get_input_names(self):
        return ['input', 'text_lengths']
        return ['input', 'text_lengths', 'vad_mask']

    def get_output_names(self):
        return ['logits']
@@ -66,14 +70,13 @@
    def get_dynamic_axes(self):
        return {
            'input': {
                0: 'batch_size',
                1: 'feats_length'
            },
            'text_lengths': {
                0: 'batch_size',
            'vad_mask': {
                2: 'feats_length1',
                3: 'feats_length2'
            },
            'logits': {
                0: 'batch_size',
                1: 'logits_length'
            },
        }