Rename 'res' in line 514 to avoid with naming conflict with line 365

liugz18

2024-07-18 d80ac2fd2df4e7fb8a28acfa512bb11472b5cc99

 funasr/models/qwen_audio/audio.py

@@ -15,10 +15,10 @@
from funasr.register import tables



def exact_div(x, y):
    assert x % y == 0
    return x // y


# hard-coded audio hyperparameters
SAMPLE_RATE = 16000
@@ -34,29 +34,36 @@
TOKENS_PER_SECOND = exact_div(SAMPLE_RATE, N_SAMPLES_PER_TOKEN)  # 20ms per audio token



def get_T_after_cnn(L_in, dilation=1):
    for (padding, kernel_size, stride) in eval("[(1,3,1)] + [(1,3,2)] "):
    for padding, kernel_size, stride in eval("[(1,3,1)] + [(1,3,2)] "):
        L_out = L_in + 2 * padding - dilation * (kernel_size - 1) - 1
        L_out = 1 + L_out // stride
        L_in = L_out
    return L_out


def load_bytesio_audio(content, sr: int = SAMPLE_RATE):
    cmd = [
        "ffmpeg",
        "-nostdin",
        "-threads", "0",
        "-i", "pipe:",
        "-f", "s16le",
        "-ac", "1",
        "-acodec", "pcm_s16le",
        "-ar", str(sr),
        "pipe:"
        "-threads",
        "0",
        "-i",
        "pipe:",
        "-f",
        "s16le",
        "-ac",
        "1",
        "-acodec",
        "pcm_s16le",
        "-ar",
        str(sr),
        "pipe:",
    ]
    p = Popen(cmd, stdin=PIPE, stdout=PIPE, stderr=PIPE, bufsize=-1)
    out, _ = p.communicate(input=content)
    return np.frombuffer(out, np.int16).flatten().astype(np.float32) / 32768.0


def load_audio(file: str, sr: int = SAMPLE_RATE):
    """
@@ -104,9 +111,7 @@
    """
    if torch.is_tensor(array):
        if array.shape[axis] > length:
            array = array.index_select(
                dim=axis, index=torch.arange(length, device=array.device)
            )
            array = array.index_select(dim=axis, index=torch.arange(length, device=array.device))

        if array.shape[axis] < length:
            pad_widths = [(0, 0)] * array.ndim
@@ -123,15 +128,14 @@

    return array


def trim(array, length: int = N_SAMPLES, *, axis: int = -1):
    """
    Pad or trim the audio array to N_SAMPLES, as expected by the encoder.
    """
    if torch.is_tensor(array):
        if array.shape[axis] > length:
            array = array.index_select(
                dim=axis, index=torch.arange(length, device=array.device)
            )
            array = array.index_select(dim=axis, index=torch.arange(length, device=array.device))
    else:
        if array.shape[axis] > length:
            array = array.take(indices=range(length), axis=axis)
@@ -151,7 +155,7 @@
    """
    assert n_mels == 80, f"Unsupported n_mels: {n_mels}"
    with np.load(
        os.path.join(os.path.dirname(__file__), "mel_filters.npz") # todo
        os.path.join(os.path.dirname(__file__), "mel_filters.npz")  # todo
        # os.path.join("assets", "mel_filters.npz")
    ) as f:
        return torch.from_numpy(f[f"mel_{n_mels}"]).to(device)
@@ -227,8 +231,6 @@
        return super().forward(x).type(x.dtype)




class Linear(nn.Linear):
    def forward(self, x: Tensor) -> Tensor:
        return F.linear(
@@ -239,9 +241,7 @@


class Conv1d(nn.Conv1d):
    def _conv_forward(
        self, x: Tensor, weight: Tensor, bias: Optional[Tensor]
    ) -> Tensor:
    def _conv_forward(self, x: Tensor, weight: Tensor, bias: Optional[Tensor]) -> Tensor:
        return super()._conv_forward(
            x, weight.to(x.dtype), None if bias is None else bias.to(x.dtype)
        )
@@ -287,9 +287,7 @@
        wv, qk = self.qkv_attention(q, k, v, mask)
        return self.out(wv), qk

    def qkv_attention(
        self, q: Tensor, k: Tensor, v: Tensor, mask: Optional[Tensor] = None
    ):
    def qkv_attention(self, q: Tensor, k: Tensor, v: Tensor, mask: Optional[Tensor] = None):
        n_batch, n_ctx, n_state = q.shape
        scale = (n_state // self.n_head) ** -0.25
        q = q.view(*q.shape[:2], self.n_head, -1).permute(0, 2, 1, 3) * scale
@@ -311,15 +309,11 @@
        self.attn = MultiHeadAttention(n_state, n_head)
        self.attn_ln = LayerNorm(n_state)

        self.cross_attn = (
            MultiHeadAttention(n_state, n_head) if cross_attention else None
        )
        self.cross_attn = MultiHeadAttention(n_state, n_head) if cross_attention else None
        self.cross_attn_ln = LayerNorm(n_state) if cross_attention else None

        n_mlp = n_state * 4
        self.mlp = nn.Sequential(
            Linear(n_state, n_mlp), nn.GELU(), Linear(n_mlp, n_state)
        )
        self.mlp = nn.Sequential(Linear(n_state, n_mlp), nn.GELU(), Linear(n_mlp, n_state))
        self.mlp_ln = LayerNorm(n_state)

    def forward(
@@ -335,19 +329,20 @@
        x = x + self.mlp(self.mlp_ln(x))
        return x


@tables.register("encoder_classes", "QwenAudioEncoder")
class QwenAudioEncoder(nn.Module):
    def __init__(
            self,
            n_mels: int,
            n_ctx: int,
            n_state: int,
            n_head: int,
            n_layer: int,
            output_dim: int = 512,
            avg_pool: bool = True,
            add_audio_bos_eos_token: bool = True,
            **kwargs
        self,
        n_mels: int,
        n_ctx: int,
        n_state: int,
        n_head: int,
        n_layer: int,
        output_dim: int = 512,
        avg_pool: bool = True,
        add_audio_bos_eos_token: bool = True,
        **kwargs,
    ):
        super().__init__()
        self.conv1 = Conv1d(n_mels, n_state, kernel_size=3, padding=1)
@@ -371,15 +366,14 @@
        self.output_dim = output_dim
        self.n_head = n_head

    def forward(self, x: Tensor, padding_mask: Tensor=None, audio_lengths: Tensor=None):
    def forward(self, x: Tensor, padding_mask: Tensor = None, audio_lengths: Tensor = None):
        """
        x : torch.Tensor, shape = (batch_size, n_mels, n_ctx)
            the mel spectrogram of the audio
        """
        x = x.to(dtype=self.conv1.weight.dtype,
                 device=self.conv1.weight.device)
        x = x.to(dtype=self.conv1.weight.dtype, device=self.conv1.weight.device)
        if audio_lengths is not None:
            input_mel_len = audio_lengths[:,0] * 2
            input_mel_len = audio_lengths[:, 0] * 2
            max_mel_len_in_batch = input_mel_len.max()
            x = x[:, :, :max_mel_len_in_batch]
        x = F.gelu(self.conv1(x))
@@ -388,34 +382,35 @@
        bsz = x.size(0)
        src_len = x.size(1)


        self.input_positional_embedding = self.positional_embedding[:src_len]
        assert x.shape[1:] == self.input_positional_embedding.shape, f"incorrect audio shape: {x.shape[1:], self.input_positional_embedding.shape}"
        assert (
            x.shape[1:] == self.input_positional_embedding.shape
        ), f"incorrect audio shape: {x.shape[1:], self.input_positional_embedding.shape}"
        x = (x + self.input_positional_embedding).to(x.dtype)
        if padding_mask is not None:
            padding_mask = padding_mask.to(dtype=self.conv1.weight.dtype,
                     device=self.conv1.weight.device)
            padding_mask = padding_mask.to(
                dtype=self.conv1.weight.dtype, device=self.conv1.weight.device
            )
            batch_src_len = padding_mask.size(1)
            x = x[:, :batch_src_len, :]
            padding_mask = padding_mask.view(
                bsz, -1, batch_src_len
            )
            padding_mask = padding_mask.view(bsz, -1, batch_src_len)
            padding_mask_ = padding_mask.all(1)
            x[padding_mask_] = 0
            key_padding_mask = padding_mask_.view(bsz, 1, 1, batch_src_len). \
                expand(-1, self.n_head, -1, -1).reshape(bsz, self.n_head, 1, batch_src_len)
            key_padding_mask = (
                padding_mask_.view(bsz, 1, 1, batch_src_len)
                .expand(-1, self.n_head, -1, -1)
                .reshape(bsz, self.n_head, 1, batch_src_len)
            )
            new_padding_mask = torch.zeros_like(key_padding_mask, dtype=x.dtype)
            padding_mask = new_padding_mask.masked_fill(key_padding_mask, float("-inf"))

        for block in self.blocks:
            x = block(x, mask=padding_mask)


        if self.avg_pooler:
            x = x.permute(0, 2, 1)
            x = self.avg_pooler(x)
            x = x.permute(0, 2, 1)


        x = self.ln_post(x)
        x = self.proj(x)
@@ -430,15 +425,16 @@
    def encode(self, input_audios: Tensor, input_audio_lengths: Tensor, audio_span_tokens: List):
        real_input_audio_lens = input_audio_lengths[:, 0].tolist()
        max_len_in_batch = max(real_input_audio_lens)
        padding_mask = torch.ones([input_audios.size(0), max_len_in_batch]).to(dtype=self.conv1.weight.dtype,
                                                                               device=self.conv1.weight.device)
        padding_mask = torch.ones([input_audios.size(0), max_len_in_batch]).to(
            dtype=self.conv1.weight.dtype, device=self.conv1.weight.device
        )
        for index in range(len(input_audios)):
            padding_mask[index, :input_audio_lengths[index][0].item()] = 0
        x, bos, eos = self(input_audios, padding_mask,input_audio_lengths)
            padding_mask[index, : input_audio_lengths[index][0].item()] = 0
        x, bos, eos = self(input_audios, padding_mask, input_audio_lengths)
        output_audios = []
        for i in range(len(audio_span_tokens)):
            audio_span = audio_span_tokens[i]
            audio = x[i][:audio_span-2]
            audio = x[i][: audio_span - 2]
            if bos is not None:
                audio = torch.concat([bos, audio, eos])
            assert len(audio) == audio_span