archs/radtts.txt

RADTTS(
  (speaker_embedding): Embedding(3, 16)
  (embedding): Embedding(185, 512)
  (flows): ModuleList(
    (0-1): 2 x FlowStep(
      (invtbl_conv): Invertible1x1ConvLUS()
      (affine_tfn): AffineTransformationLayer(
        (affine_param_predictor): WN(
          (in_layers): ModuleList(
            (0): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(2,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
            (1): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(4,), dilation=(2,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
            (2): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(8,), dilation=(4,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
            (3): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(16,), dilation=(8,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
          )
          (res_skip_layers): ModuleList(
            (0-3): 4 x ParametrizedConv1d(
              1024, 1024, kernel_size=(1,), stride=(1,)
              (parametrizations): ModuleDict(
                (weight): ParametrizationList(
                  (0): _WeightNorm()
                )
              )
            )
          )
          (start): ParametrizedConv1d(
            1120, 1024, kernel_size=(1,), stride=(1,)
            (parametrizations): ModuleDict(
              (weight): ParametrizationList(
                (0): _WeightNorm()
              )
            )
          )
          (softplus): Softplus(beta=1.0, threshold=20.0)
          (end): Conv1d(1024, 160, kernel_size=(1,), stride=(1,))
        )
      )
    )
    (2-3): 2 x FlowStep(
      (invtbl_conv): Invertible1x1ConvLUS()
      (affine_tfn): AffineTransformationLayer(
        (affine_param_predictor): WN(
          (in_layers): ModuleList(
            (0): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(2,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
            (1): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(4,), dilation=(2,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
            (2): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(8,), dilation=(4,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
            (3): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(16,), dilation=(8,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
          )
          (res_skip_layers): ModuleList(
            (0-3): 4 x ParametrizedConv1d(
              1024, 1024, kernel_size=(1,), stride=(1,)
              (parametrizations): ModuleDict(
                (weight): ParametrizationList(
                  (0): _WeightNorm()
                )
              )
            )
          )
          (start): ParametrizedConv1d(
            1119, 1024, kernel_size=(1,), stride=(1,)
            (parametrizations): ModuleDict(
              (weight): ParametrizationList(
                (0): _WeightNorm()
              )
            )
          )
          (softplus): Softplus(beta=1.0, threshold=20.0)
          (end): Conv1d(1024, 158, kernel_size=(1,), stride=(1,))
        )
      )
    )
    (4-5): 2 x FlowStep(
      (invtbl_conv): Invertible1x1ConvLUS()
      (affine_tfn): AffineTransformationLayer(
        (affine_param_predictor): WN(
          (in_layers): ModuleList(
            (0): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(2,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
            (1): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(4,), dilation=(2,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
            (2): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(8,), dilation=(4,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
            (3): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(16,), dilation=(8,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
          )
          (res_skip_layers): ModuleList(
            (0-3): 4 x ParametrizedConv1d(
              1024, 1024, kernel_size=(1,), stride=(1,)
              (parametrizations): ModuleDict(
                (weight): ParametrizationList(
                  (0): _WeightNorm()
                )
              )
            )
          )
          (start): ParametrizedConv1d(
            1118, 1024, kernel_size=(1,), stride=(1,)
            (parametrizations): ModuleDict(
              (weight): ParametrizationList(
                (0): _WeightNorm()
              )
            )
          )
          (softplus): Softplus(beta=1.0, threshold=20.0)
          (end): Conv1d(1024, 156, kernel_size=(1,), stride=(1,))
        )
      )
    )
    (6-7): 2 x FlowStep(
      (invtbl_conv): Invertible1x1ConvLUS()
      (affine_tfn): AffineTransformationLayer(
        (affine_param_predictor): WN(
          (in_layers): ModuleList(
            (0): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(2,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
            (1): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(4,), dilation=(2,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
            (2): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(8,), dilation=(4,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
            (3): ConvNorm(
              (conv): ParametrizedPartialConv1d(
                1024, 1024, kernel_size=(5,), stride=(1,), padding=(16,), dilation=(8,)
                (parametrizations): ModuleDict(
                  (weight): ParametrizationList(
                    (0): _WeightNorm()
                  )
                )
              )
            )
          )
          (res_skip_layers): ModuleList(
            (0-3): 4 x ParametrizedConv1d(
              1024, 1024, kernel_size=(1,), stride=(1,)
              (parametrizations): ModuleDict(
                (weight): ParametrizationList(
                  (0): _WeightNorm()
                )
              )
            )
          )
          (start): ParametrizedConv1d(
            1117, 1024, kernel_size=(1,), stride=(1,)
            (parametrizations): ModuleDict(
              (weight): ParametrizationList(
                (0): _WeightNorm()
              )
            )
          )
          (softplus): Softplus(beta=1.0, threshold=20.0)
          (end): Conv1d(1024, 154, kernel_size=(1,), stride=(1,))
        )
      )
    )
  )
  (encoder): Encoder(
    (convolutions): ModuleList(
      (0-2): 3 x Sequential(
        (0): ConvNorm(
          (conv): PartialConv1d(512, 512, kernel_size=(5,), stride=(1,), padding=(2,))
        )
        (1): InstanceNorm1d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=False)
      )
    )
    (lstm): ParametrizedLSTM(
      512, 256, batch_first=True, bidirectional=True
      (parametrizations): ModuleDict(
        (weight_hh_l0): ParametrizationList(
          (0): _SpectralNorm()
        )
        (weight_hh_l0_reverse): ParametrizationList(
          (0): _SpectralNorm()
        )
      )
    )
  )
  (length_regulator): LengthRegulator()
  (attention): ConvAttention(
    (softmax): Softmax(dim=3)
    (log_softmax): LogSoftmax(dim=3)
    (key_proj): Sequential(
      (0): ConvNorm(
        (conv): Conv1d(512, 1024, kernel_size=(3,), stride=(1,), padding=(1,))
      )
      (1): ReLU()
      (2): ConvNorm(
        (conv): Conv1d(1024, 80, kernel_size=(1,), stride=(1,))
      )
    )
    (query_proj): Sequential(
      (0): ConvNorm(
        (conv): Conv1d(80, 160, kernel_size=(3,), stride=(1,), padding=(1,))
      )
      (1): ReLU()
      (2): ConvNorm(
        (conv): Conv1d(160, 80, kernel_size=(1,), stride=(1,))
      )
      (3): ReLU()
      (4): ConvNorm(
        (conv): Conv1d(80, 80, kernel_size=(1,), stride=(1,))
      )
    )
  )
  (context_lstm): ParametrizedLSTM(
    1044, 520, batch_first=True, bidirectional=True
    (parametrizations): ModuleDict(
      (weight_hh_l0): ParametrizationList(
        (0): _SpectralNorm()
      )
      (weight_hh_l0_reverse): ParametrizationList(
        (0): _SpectralNorm()
      )
    )
  )
  (unfold): Unfold(kernel_size=(2, 1), dilation=1, padding=0, stride=2)
  (dur_pred_layer): DAP(
    (bottleneck_layer): BottleneckLayerLayer(
      (projection_fn): ConvNorm(
        (conv): ParametrizedConv1d(
          512, 32, kernel_size=(3,), stride=(1,), padding=(1,)
          (parametrizations): ModuleDict(
            (weight): ParametrizationList(
              (0): _WeightNorm()
            )
          )
        )
      )
      (non_linearity): ReLU()
    )
    (feat_pred_fn): ConvLSTMLinear(
      (dropout): Dropout(p=0.25, inplace=False)
      (convolutions): ModuleList(
        (0): ParametrizedConv1d(
          48, 256, kernel_size=(3,), stride=(1,), padding=(1,)
          (parametrizations): ModuleDict(
            (weight): ParametrizationList(
              (0): _WeightNorm()
            )
          )
        )
        (1): ParametrizedConv1d(
          256, 256, kernel_size=(3,), stride=(1,), padding=(1,)
          (parametrizations): ModuleDict(
            (weight): ParametrizationList(
              (0): _WeightNorm()
            )
          )
        )
      )
      (bilstm): ParametrizedLSTM(
        256, 128, batch_first=True, bidirectional=True
        (parametrizations): ModuleDict(
          (weight_hh_l0): ParametrizationList(
            (0): _SpectralNorm()
          )
          (weight_hh_l0_reverse): ParametrizationList(
            (0): _SpectralNorm()
          )
        )
      )
      (dense): Linear(in_features=256, out_features=1, bias=True)
    )
  )
  (unvoiced_bias_module): Sequential(
    (0): LinearNorm(
      (linear_layer): Linear(in_features=512, out_features=1, bias=True)
    )
    (1): ReLU()
  )
  (v_pred_module): DAP(
    (bottleneck_layer): BottleneckLayerLayer(
      (projection_fn): ConvNorm(
        (conv): ParametrizedConv1d(
          512, 32, kernel_size=(3,), stride=(1,), padding=(1,)
          (parametrizations): ModuleDict(
            (weight): ParametrizationList(
              (0): _WeightNorm()
            )
          )
        )
      )
      (non_linearity): ReLU()
    )
    (feat_pred_fn): ConvLSTMLinear(
      (dropout): Dropout(p=0.5, inplace=False)
      (convolutions): ModuleList(
        (0): ParametrizedConv1d(
          48, 256, kernel_size=(3,), stride=(1,), padding=(1,)
          (parametrizations): ModuleDict(
            (weight): ParametrizationList(
              (0): _WeightNorm()
            )
          )
        )
        (1): ParametrizedConv1d(
          256, 256, kernel_size=(3,), stride=(1,), padding=(1,)
          (parametrizations): ModuleDict(
            (weight): ParametrizationList(
              (0): _WeightNorm()
            )
          )
        )
      )
      (dense): Linear(in_features=256, out_features=1, bias=True)
    )
  )
  (v_embeddings): Embedding(4, 512)
  (f0_pred_module): DAP(
    (bottleneck_layer): BottleneckLayerLayer(
      (projection_fn): ConvNorm(
        (conv): ParametrizedConv1d(
          512, 32, kernel_size=(3,), stride=(1,), padding=(1,)
          (parametrizations): ModuleDict(
            (weight): ParametrizationList(
              (0): _WeightNorm()
            )
          )
        )
      )
      (non_linearity): ReLU()
    )
    (feat_pred_fn): ConvLSTMLinear(
      (dropout): Dropout(p=0.5, inplace=False)
      (convolutions): ModuleList(
        (0): ParametrizedConv1d(
          48, 256, kernel_size=(11,), stride=(1,), padding=(5,)
          (parametrizations): ModuleDict(
            (weight): ParametrizationList(
              (0): _WeightNorm()
            )
          )
        )
        (1): ParametrizedConv1d(
          256, 256, kernel_size=(11,), stride=(1,), padding=(5,)
          (parametrizations): ModuleDict(
            (weight): ParametrizationList(
              (0): _WeightNorm()
            )
          )
        )
      )
      (bilstm): ParametrizedLSTM(
        256, 128, batch_first=True, bidirectional=True
        (parametrizations): ModuleDict(
          (weight_hh_l0): ParametrizationList(
            (0): _SpectralNorm()
          )
          (weight_hh_l0_reverse): ParametrizationList(
            (0): _SpectralNorm()
          )
        )
      )
      (dense): Linear(in_features=256, out_features=1, bias=True)
    )
  )
  (energy_pred_module): DAP(
    (bottleneck_layer): BottleneckLayerLayer(
      (projection_fn): ConvNorm(
        (conv): ParametrizedConv1d(
          512, 32, kernel_size=(3,), stride=(1,), padding=(1,)
          (parametrizations): ModuleDict(
            (weight): ParametrizationList(
              (0): _WeightNorm()
            )
          )
        )
      )
      (non_linearity): ReLU()
    )
    (feat_pred_fn): ConvLSTMLinear(
      (dropout): Dropout(p=0.25, inplace=False)
      (convolutions): ModuleList(
        (0): ParametrizedConv1d(
          48, 256, kernel_size=(3,), stride=(1,), padding=(1,)
          (parametrizations): ModuleDict(
            (weight): ParametrizationList(
              (0): _WeightNorm()
            )
          )
        )
        (1): ParametrizedConv1d(
          256, 256, kernel_size=(3,), stride=(1,), padding=(1,)
          (parametrizations): ModuleDict(
            (weight): ParametrizationList(
              (0): _WeightNorm()
            )
          )
        )
      )
      (bilstm): ParametrizedLSTM(
        256, 128, batch_first=True, bidirectional=True
        (parametrizations): ModuleDict(
          (weight_hh_l0): ParametrizationList(
            (0): _SpectralNorm()
          )
          (weight_hh_l0_reverse): ParametrizationList(
            (0): _SpectralNorm()
          )
        )
      )
      (dense): Linear(in_features=256, out_features=1, bias=True)
    )
  )
)