Skip to content

Instantly share code, notes, and snippets.

@albertz
Created November 26, 2020 17:22
Show Gist options
  • Save albertz/7f3834965d69472f6c7246132f5d186e to your computer and use it in GitHub Desktop.
Save albertz/7f3834965d69472f6c7246132f5d186e to your computer and use it in GitHub Desktop.
PyTorch to RETURNN, Parallel WaveGAN example, full log output
/usr/local/bin/python3 "/Users/az/Library/Application Support/JetBrains/Toolbox/apps/PyCharm-C/ch-0/202.7660.27/PyCharm CE.app/Contents/plugins/python-ce/helpers/pydev/pydevd.py" --multiproc --qt-support=auto --client 127.0.0.1 --port 57798 --file /Users/az/Programmierung/import-parallel-wavegan/pytorch_to_returnn.py --pwg_config mb_melgan.v2.yaml --pwg_checkpoint mb_melgan_models/checkpoint-1000000steps.pkl --features data/features.npy
pydev debugger: process 58079 is connecting
Connected to pydev debugger (build 202.7660.27)
2020-11-26 18:21:44.088082: I tensorflow/core/platform/cpu_feature_guard.cc:143] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2020-11-26 18:21:44.101822: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x7fabda13afa0 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
2020-11-26 18:21:44.101846: I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): Host, Default Version
CUDA_VISIBLE_DEVICES is not set.
Collecting TensorFlow device list...
Local devices available to TensorFlow:
1/2: name: "/device:CPU:0"
device_type: "CPU"
memory_limit: 268435456
locality {
}
incarnation: 3807773853604125641
2/2: name: "/device:XLA_CPU:0"
device_type: "XLA_CPU"
memory_limit: 17179869184
locality {
}
incarnation: 8874893949683666076
physical_device_desc: "device: XLA_CPU device"
Feature shape: (80, 80)
>>> Running with standard reference imports...
>>> Running with wrapped imports, wrapping original PyTorch...
*** register sys.meta_path for ctx <WrapCtx 'pytorch_to_returnn.import_wrapper._torch_traced'>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.from_numpy' -> <class 'pytorch_to_returnn.import_wrapper.base_wrappers.function._VariableFunctionsClass.from_numpy'>
*** func call pytorch_to_returnn.import_wrapper._torch_traced.torch.from_numpy(...)
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.parallel_wavegan>
*** <WrapCtx 'pytorch_to_returnn.import_wrapper._torch_traced'> extend by mod 'parallel_wavegan'
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.parallel_wavegan.models>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.parallel_wavegan.models.melgan>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.parallel_wavegan.layers>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.parallel_wavegan.layers.causal_conv>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn' -> <WrappedIndirectModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.Module' -> <class 'pytorch_to_returnn.import_wrapper.torch_wrappers.module.WrappedModuleBase'>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.parallel_wavegan.layers.pqmf>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.parallel_wavegan.layers.residual_block>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.module.Module' -> <class 'pytorch_to_returnn.import_wrapper.torch_wrappers.module.WrappedModuleBase'>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.linear>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.Tensor' -> <class 'pytorch_to_returnn.import_wrapper.torch_wrappers.tensor.WrappedTorchTensor'>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.parameter.Parameter' -> <class 'pytorch_to_returnn.import_wrapper.torch_wrappers.parameter.WrappedTorchParameter'>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.init' -> <WrappedIndirectModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.init>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.conv>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.utils>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch._six.container_abcs'
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.common_types._size_1_t'
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.common_types._size_2_t'
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.common_types._size_3_t'
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.activation>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.init.xavier_uniform_' -> <function xavier_uniform_ at 0x10b91d440>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.init.constant_' -> <function constant_ at 0x10b91d4d0>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.init.xavier_normal_' -> <function xavier_normal_ at 0x10b91d560>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.loss>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.distance>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn._reduction' -> <WrappedIndirectModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn._reduction>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.container>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch._jit_internal._copy_to_script_wrapper' -> <function _copy_to_script_wrapper at 0x1552d0560>
*** func call pytorch_to_returnn.import_wrapper._torch_traced.torch._jit_internal._copy_to_script_wrapper(...)
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.pooling>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.common_types._size_any_t'
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.common_types._ratio_3_t'
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.common_types._ratio_2_t'
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.batchnorm>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules._functions>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.autograd.function.Function'
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.instancenorm>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.normalization>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.Size'
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.dropout>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.padding>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.common_types._size_4_t'
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.common_types._size_6_t'
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.sparse>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.rnn>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.utils.rnn.PackedSequence' -> <class 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.utils.rnn.PackedSequence'>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch._VF' -> <WrappedIndirectModule pytorch_to_returnn.import_wrapper._torch_traced.torch._VF>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch._VF.rnn_tanh' -> <class 'pytorch_to_returnn.import_wrapper.base_wrappers.function._VariableFunctionsClass.rnn_tanh'>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch._VF.rnn_relu' -> <class 'pytorch_to_returnn.import_wrapper.base_wrappers.function._VariableFunctionsClass.rnn_relu'>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch._jit_internal._overload_method' -> <function _overload_method at 0x1553a8ef0>
*** func call pytorch_to_returnn.import_wrapper._torch_traced.torch._jit_internal._overload_method(...)
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.pixelshuffle>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.upsampling>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.common_types._ratio_any_t'
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.fold>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.adaptive>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.functional.log_softmax' -> <function log_softmax at 0x15a90fd40>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.transformer>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.flatten>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.Conv1d' -> <class 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.conv.Conv1d'>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.parallel_wavegan.layers.residual_stack>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.parallel_wavegan.layers.upsample>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.Conv2d' -> <class 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.conv.Conv2d'>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_traced.parallel_wavegan.models.parallel_wavegan>
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.device'
*** torch module create pytorch_to_returnn.import_wrapper._torch_traced.parallel_wavegan.models.melgan.MelGANGenerator(...)
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.ReflectionPad1d' -> <class 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.padding.ReflectionPad1d'>
*** torch module create pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.padding.ReflectionPad1d(...)
*** torch module create pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.conv.Conv1d(...)
**** torch tensor func __get__
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.init.kaiming_uniform_' -> <function kaiming_uniform_ at 0x154d70e60>
*** func call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.init.kaiming_uniform_(...)
**** torch tensor func dim
**** torch tensor func size
**** torch tensor func __getitem__
**** torch tensor func uniform_
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.init._calculate_fan_in_and_fan_out' -> <function _calculate_fan_in_and_fan_out at 0x155360b90>
*** func call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.init._calculate_fan_in_and_fan_out(...)
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.init.uniform_' -> <function uniform_ at 0x155360c20>
*** func call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.init.uniform_(...)
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.LeakyReLU' -> <class 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.activation.LeakyReLU'>
*** torch module create pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.activation.LeakyReLU(...)
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.ConvTranspose1d' -> <class 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.conv.ConvTranspose1d'>
*** torch module create pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.conv.ConvTranspose1d(...)
*** torch module create pytorch_to_returnn.import_wrapper._torch_traced.parallel_wavegan.layers.residual_stack.ResidualStack(...)
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.Sequential' -> <class 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.container.Sequential'>
*** torch module create pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.container.Sequential(...)
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.Tanh' -> <class 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.activation.Tanh'>
*** torch module create pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.activation.Tanh(...)
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.utils.weight_norm' -> <function weight_norm at 0x1553c1950>
*** func call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.utils.weight_norm(...)
**** torch tensor func norm_except_dim
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.load' -> <function load at 0x1553c18c0>
*** func call pytorch_to_returnn.import_wrapper._torch_traced.torch.load(...)
**** torch tensor func copy_
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.utils.remove_weight_norm' -> <function remove_weight_norm at 0x1553c1f80>
*** func call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.utils.remove_weight_norm(...)
**** torch tensor func is_floating_point
**** torch tensor func to
**** torch tensor func _has_compatible_shallow_copy_type
**** torch tensor func __set__
*** torch module create pytorch_to_returnn.import_wrapper._torch_traced.parallel_wavegan.layers.pqmf.PQMF(...)
**** torch tensor func float
**** torch tensor func unsqueeze
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.zeros' -> <class 'pytorch_to_returnn.import_wrapper.base_wrappers.function._VariableFunctionsClass.zeros'>
*** func call pytorch_to_returnn.import_wrapper._torch_traced.torch.zeros(...)
**** torch tensor func __setitem__
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.ConstantPad1d' -> <class 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.padding.ConstantPad1d'>
*** torch module create pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.padding.ConstantPad1d(...)
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.no_grad' -> <class 'pytorch_to_returnn.import_wrapper._torch_traced.torch.autograd.grad_mode.no_grad'>
*** WrappedClass pytorch_to_returnn.import_wrapper._torch_traced.torch.no_grad(...)
*** torch module call pytorch_to_returnn.import_wrapper._torch_traced.parallel_wavegan.models.melgan.MelGANGenerator(...)(...)
**** torch tensor func __hash__
*** torch module call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.container.Sequential(...)(...)
*** torch module call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.padding.ReflectionPad1d(...)(...)
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.functional.pad' -> <function _pad at 0x1552f08c0>
*** func call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.functional.pad(...)
**** torch tensor func _pad
*** torch module call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.conv.Conv1d(...)(...)
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.functional.conv1d' -> <class 'pytorch_to_returnn.import_wrapper.base_wrappers.function._VariableFunctionsClass.conv1d'>
*** func call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.functional.conv1d(...)
**** torch tensor func conv1d
*** torch module call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.activation.LeakyReLU(...)(...)
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.functional.leaky_relu' -> <function leaky_relu at 0x1552f0b90>
*** func call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.functional.leaky_relu(...)
**** torch tensor func leaky_relu
*** torch module call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.conv.ConvTranspose1d(...)(...)
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.functional.conv_transpose1d' -> <class 'pytorch_to_returnn.import_wrapper.base_wrappers.function._VariableFunctionsClass.conv_transpose1d'>
*** func call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.functional.conv_transpose1d(...)
**** torch tensor func conv_transpose1d
*** torch module call pytorch_to_returnn.import_wrapper._torch_traced.parallel_wavegan.layers.residual_stack.ResidualStack(...)(...)
**** torch tensor func add
*** torch module call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.activation.Tanh(...)(...)
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.tanh' -> <class 'pytorch_to_returnn.import_wrapper.base_wrappers.function._VariableFunctionsClass.tanh'>
*** func call pytorch_to_returnn.import_wrapper._torch_traced.torch.tanh(...)
**** torch tensor func tanh
**** torch tensor func mul
*** torch module call pytorch_to_returnn.import_wrapper._torch_traced.torch.nn.modules.padding.ConstantPad1d(...)(...)
*** indirect getattr 'pytorch_to_returnn.import_wrapper._torch_traced.torch.no_grad(...).prev'
**** torch tensor func cpu
**** torch tensor func numpy
>>>> Module naming hierarchy:
.tmp_root: (hidden)
melgan: <ModuleEntry Sequential(...)> -> ...
layer0: <ModuleEntry ReflectionPad1d((3, 3))> -> ...
layer1: <ModuleEntry Conv1d(80, 384, kernel_size=(7,), stride=(1,))> -> ...
layer2: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer3: <ModuleEntry ConvTranspose1d(384, 192, kernel_size=(10,), stride=(5,), padding=(3,), output_padding=(1,))> -> ...
layer4: <ModuleEntry ResidualStack(...)> -> ...
stack: <ModuleEntry Sequential(...)> -> ...
layer0: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer1: <ModuleEntry ReflectionPad1d((1, 1))> -> ...
layer2: <ModuleEntry Conv1d(192, 192, kernel_size=(3,), stride=(1,))> -> ...
layer3: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer4: <ModuleEntry Conv1d(192, 192, kernel_size=(1,), stride=(1,))> -> ...
skip_layer: <ModuleEntry Conv1d(192, 192, kernel_size=(1,), stride=(1,))> -> ...
layer5: <ModuleEntry ResidualStack(...)> -> ...
stack: <ModuleEntry Sequential(...)> -> ...
layer0: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer1: <ModuleEntry ReflectionPad1d((3, 3))> -> ...
layer2: <ModuleEntry Conv1d(192, 192, kernel_size=(3,), stride=(1,), dilation=(3,))> -> ...
layer3: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer4: <ModuleEntry Conv1d(192, 192, kernel_size=(1,), stride=(1,))> -> ...
skip_layer: <ModuleEntry Conv1d(192, 192, kernel_size=(1,), stride=(1,))> -> ...
layer6: <ModuleEntry ResidualStack(...)> -> ...
stack: <ModuleEntry Sequential(...)> -> ...
layer0: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer1: <ModuleEntry ReflectionPad1d((9, 9))> -> ...
layer2: <ModuleEntry Conv1d(192, 192, kernel_size=(3,), stride=(1,), dilation=(9,))> -> ...
layer3: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer4: <ModuleEntry Conv1d(192, 192, kernel_size=(1,), stride=(1,))> -> ...
skip_layer: <ModuleEntry Conv1d(192, 192, kernel_size=(1,), stride=(1,))> -> ...
layer7: <ModuleEntry ResidualStack(...)> -> ...
stack: <ModuleEntry Sequential(...)> -> ...
layer0: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer1: <ModuleEntry ReflectionPad1d((27, 27))> -> ...
layer2: <ModuleEntry Conv1d(192, 192, kernel_size=(3,), stride=(1,), dilation=(27,))> -> ...
layer3: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer4: <ModuleEntry Conv1d(192, 192, kernel_size=(1,), stride=(1,))> -> ...
skip_layer: <ModuleEntry Conv1d(192, 192, kernel_size=(1,), stride=(1,))> -> ...
layer8: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer9: <ModuleEntry ConvTranspose1d(192, 96, kernel_size=(10,), stride=(5,), padding=(3,), output_padding=(1,))> -> ...
layer10: <ModuleEntry ResidualStack(...)> -> ...
stack: <ModuleEntry Sequential(...)> -> ...
layer0: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer1: <ModuleEntry ReflectionPad1d((1, 1))> -> ...
layer2: <ModuleEntry Conv1d(96, 96, kernel_size=(3,), stride=(1,))> -> ...
layer3: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer4: <ModuleEntry Conv1d(96, 96, kernel_size=(1,), stride=(1,))> -> ...
skip_layer: <ModuleEntry Conv1d(96, 96, kernel_size=(1,), stride=(1,))> -> ...
layer11: <ModuleEntry ResidualStack(...)> -> ...
stack: <ModuleEntry Sequential(...)> -> ...
layer0: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer1: <ModuleEntry ReflectionPad1d((3, 3))> -> ...
layer2: <ModuleEntry Conv1d(96, 96, kernel_size=(3,), stride=(1,), dilation=(3,))> -> ...
layer3: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer4: <ModuleEntry Conv1d(96, 96, kernel_size=(1,), stride=(1,))> -> ...
skip_layer: <ModuleEntry Conv1d(96, 96, kernel_size=(1,), stride=(1,))> -> ...
layer12: <ModuleEntry ResidualStack(...)> -> ...
stack: <ModuleEntry Sequential(...)> -> ...
layer0: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer1: <ModuleEntry ReflectionPad1d((9, 9))> -> ...
layer2: <ModuleEntry Conv1d(96, 96, kernel_size=(3,), stride=(1,), dilation=(9,))> -> ...
layer3: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer4: <ModuleEntry Conv1d(96, 96, kernel_size=(1,), stride=(1,))> -> ...
skip_layer: <ModuleEntry Conv1d(96, 96, kernel_size=(1,), stride=(1,))> -> ...
layer13: <ModuleEntry ResidualStack(...)> -> ...
stack: <ModuleEntry Sequential(...)> -> ...
layer0: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer1: <ModuleEntry ReflectionPad1d((27, 27))> -> ...
layer2: <ModuleEntry Conv1d(96, 96, kernel_size=(3,), stride=(1,), dilation=(27,))> -> ...
layer3: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer4: <ModuleEntry Conv1d(96, 96, kernel_size=(1,), stride=(1,))> -> ...
skip_layer: <ModuleEntry Conv1d(96, 96, kernel_size=(1,), stride=(1,))> -> ...
layer14: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer15: <ModuleEntry ConvTranspose1d(96, 48, kernel_size=(4,), stride=(2,), padding=(1,))> -> ...
layer16: <ModuleEntry ResidualStack(...)> -> ...
stack: <ModuleEntry Sequential(...)> -> ...
layer0: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer1: <ModuleEntry ReflectionPad1d((1, 1))> -> ...
layer2: <ModuleEntry Conv1d(48, 48, kernel_size=(3,), stride=(1,))> -> ...
layer3: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer4: <ModuleEntry Conv1d(48, 48, kernel_size=(1,), stride=(1,))> -> ...
skip_layer: <ModuleEntry Conv1d(48, 48, kernel_size=(1,), stride=(1,))> -> ...
layer17: <ModuleEntry ResidualStack(...)> -> ...
stack: <ModuleEntry Sequential(...)> -> ...
layer0: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer1: <ModuleEntry ReflectionPad1d((3, 3))> -> ...
layer2: <ModuleEntry Conv1d(48, 48, kernel_size=(3,), stride=(1,), dilation=(3,))> -> ...
layer3: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer4: <ModuleEntry Conv1d(48, 48, kernel_size=(1,), stride=(1,))> -> ...
skip_layer: <ModuleEntry Conv1d(48, 48, kernel_size=(1,), stride=(1,))> -> ...
layer18: <ModuleEntry ResidualStack(...)> -> ...
stack: <ModuleEntry Sequential(...)> -> ...
layer0: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer1: <ModuleEntry ReflectionPad1d((9, 9))> -> ...
layer2: <ModuleEntry Conv1d(48, 48, kernel_size=(3,), stride=(1,), dilation=(9,))> -> ...
layer3: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer4: <ModuleEntry Conv1d(48, 48, kernel_size=(1,), stride=(1,))> -> ...
skip_layer: <ModuleEntry Conv1d(48, 48, kernel_size=(1,), stride=(1,))> -> ...
layer19: <ModuleEntry ResidualStack(...)> -> ...
stack: <ModuleEntry Sequential(...)> -> ...
layer0: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer1: <ModuleEntry ReflectionPad1d((27, 27))> -> ...
layer2: <ModuleEntry Conv1d(48, 48, kernel_size=(3,), stride=(1,), dilation=(27,))> -> ...
layer3: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer4: <ModuleEntry Conv1d(48, 48, kernel_size=(1,), stride=(1,))> -> ...
skip_layer: <ModuleEntry Conv1d(48, 48, kernel_size=(1,), stride=(1,))> -> ...
layer20: <ModuleEntry LeakyReLU(negative_slope=0.2)> -> ...
layer21: <ModuleEntry ReflectionPad1d((3, 3))> -> ...
layer22: <ModuleEntry Conv1d(48, 4, kernel_size=(7,), stride=(1,))> -> ...
layer23: <ModuleEntry Tanh()> -> ...
pad_fn: <ModuleEntry ConstantPad1d(padding=(31, 31), value=0.0)> -> ...
>>>> Root module calls:
{
'melgan': <CallEntry #1 <ModuleEntry Sequential(...)>>,
'pad_fn': <CallEntry #0 <ModuleEntry ConstantPad1d(padding=(31, 31), value=0.0)>>
}
>>>> Modules with params:
{
'melgan.layer1': Conv1d(80, 384, kernel_size=(7,), stride=(1,)),
'melgan.layer3': ConvTranspose1d(384, 192, kernel_size=(10,), stride=(5,), padding=(3,), output_padding=(1,)),
'melgan.layer4.stack.layer2': Conv1d(192, 192, kernel_size=(3,), stride=(1,)),
'melgan.layer4.stack.layer4': Conv1d(192, 192, kernel_size=(1,), stride=(1,)),
'melgan.layer4.skip_layer': Conv1d(192, 192, kernel_size=(1,), stride=(1,)),
'melgan.layer5.stack.layer2': Conv1d(192, 192, kernel_size=(3,), stride=(1,), dilation=(3,)),
'melgan.layer5.stack.layer4': Conv1d(192, 192, kernel_size=(1,), stride=(1,)),
'melgan.layer5.skip_layer': Conv1d(192, 192, kernel_size=(1,), stride=(1,)),
'melgan.layer6.stack.layer2': Conv1d(192, 192, kernel_size=(3,), stride=(1,), dilation=(9,)),
'melgan.layer6.stack.layer4': Conv1d(192, 192, kernel_size=(1,), stride=(1,)),
'melgan.layer6.skip_layer': Conv1d(192, 192, kernel_size=(1,), stride=(1,)),
'melgan.layer7.stack.layer2': Conv1d(192, 192, kernel_size=(3,), stride=(1,), dilation=(27,)),
'melgan.layer7.stack.layer4': Conv1d(192, 192, kernel_size=(1,), stride=(1,)),
'melgan.layer7.skip_layer': Conv1d(192, 192, kernel_size=(1,), stride=(1,)),
'melgan.layer9': ConvTranspose1d(192, 96, kernel_size=(10,), stride=(5,), padding=(3,), output_padding=(1,)),
'melgan.layer10.stack.layer2': Conv1d(96, 96, kernel_size=(3,), stride=(1,)),
'melgan.layer10.stack.layer4': Conv1d(96, 96, kernel_size=(1,), stride=(1,)),
'melgan.layer10.skip_layer': Conv1d(96, 96, kernel_size=(1,), stride=(1,)),
'melgan.layer11.stack.layer2': Conv1d(96, 96, kernel_size=(3,), stride=(1,), dilation=(3,)),
'melgan.layer11.stack.layer4': Conv1d(96, 96, kernel_size=(1,), stride=(1,)),
'melgan.layer11.skip_layer': Conv1d(96, 96, kernel_size=(1,), stride=(1,)),
'melgan.layer12.stack.layer2': Conv1d(96, 96, kernel_size=(3,), stride=(1,), dilation=(9,)),
'melgan.layer12.stack.layer4': Conv1d(96, 96, kernel_size=(1,), stride=(1,)),
'melgan.layer12.skip_layer': Conv1d(96, 96, kernel_size=(1,), stride=(1,)),
'melgan.layer13.stack.layer2': Conv1d(96, 96, kernel_size=(3,), stride=(1,), dilation=(27,)),
'melgan.layer13.stack.layer4': Conv1d(96, 96, kernel_size=(1,), stride=(1,)),
'melgan.layer13.skip_layer': Conv1d(96, 96, kernel_size=(1,), stride=(1,)),
'melgan.layer15': ConvTranspose1d(96, 48, kernel_size=(4,), stride=(2,), padding=(1,)),
'melgan.layer16.stack.layer2': Conv1d(48, 48, kernel_size=(3,), stride=(1,)),
'melgan.layer16.stack.layer4': Conv1d(48, 48, kernel_size=(1,), stride=(1,)),
'melgan.layer16.skip_layer': Conv1d(48, 48, kernel_size=(1,), stride=(1,)),
'melgan.layer17.stack.layer2': Conv1d(48, 48, kernel_size=(3,), stride=(1,), dilation=(3,)),
'melgan.layer17.stack.layer4': Conv1d(48, 48, kernel_size=(1,), stride=(1,)),
'melgan.layer17.skip_layer': Conv1d(48, 48, kernel_size=(1,), stride=(1,)),
'melgan.layer18.stack.layer2': Conv1d(48, 48, kernel_size=(3,), stride=(1,), dilation=(9,)),
'melgan.layer18.stack.layer4': Conv1d(48, 48, kernel_size=(1,), stride=(1,)),
'melgan.layer18.skip_layer': Conv1d(48, 48, kernel_size=(1,), stride=(1,)),
'melgan.layer19.stack.layer2': Conv1d(48, 48, kernel_size=(3,), stride=(1,), dilation=(27,)),
'melgan.layer19.stack.layer4': Conv1d(48, 48, kernel_size=(1,), stride=(1,)),
'melgan.layer19.skip_layer': Conv1d(48, 48, kernel_size=(1,), stride=(1,)),
'melgan.layer22': Conv1d(48, 4, kernel_size=(7,), stride=(1,))
}
>>>> Looks good!
>>> Running with wrapped Torch import, wrapping replacement for PyTorch...
*** register sys.meta_path for ctx <WrapCtx 'pytorch_to_returnn.import_wrapper._torch_returnn'>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_returnn.parallel_wavegan>
*** <WrapCtx 'pytorch_to_returnn.import_wrapper._torch_returnn'> extend by mod 'parallel_wavegan'
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_returnn.parallel_wavegan.models>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_returnn.parallel_wavegan.models.melgan>
WARNING:tensorflow:From /Users/az/Programmierung/import-parallel-wavegan/returnn/tf/network.py:352: calling Zeros.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.
Instructions for updating:
Call initializer instance with the dtype argument instead of passing it to the constructor
WARNING:tensorflow:From /Users/az/Library/Python/3.7/lib/python/site-packages/tensorflow/python/ops/resource_variable_ops.py:1666: calling BaseResourceVariable.__init__ (from tensorflow.python.ops.resource_variable_ops) with constraint is deprecated and will be removed in a future version.
Instructions for updating:
If using Keras pass *_constraint arguments to layers.
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_returnn.parallel_wavegan.layers>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_returnn.parallel_wavegan.layers.causal_conv>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_returnn.parallel_wavegan.layers.pqmf>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_returnn.parallel_wavegan.layers.residual_block>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_returnn.parallel_wavegan.layers.residual_stack>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_returnn.parallel_wavegan.layers.upsample>
*** exec mod <WrappedSourceModule pytorch_to_returnn.import_wrapper._torch_returnn.parallel_wavegan.models.parallel_wavegan>
*** root/.tmp_root:subnet/melgan:subnet/'layer1_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (384, 80, 7)}
*** root/.tmp_root:subnet/melgan:subnet/'layer1_weight' VariableLayer output: [384,80,F|7]
*** root/.tmp_root:subnet/melgan:subnet/'layer1_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer1_weight'}
*** root/.tmp_root:subnet/melgan:subnet/'layer1_Norm' MathNormLayer output: [384,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer1_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (384, 80, 7)}
*** root/.tmp_root:subnet/melgan:subnet/'layer1_weight_v' VariableLayer output: [384,80,F|7]
*** root/.tmp_root:subnet/melgan:subnet/'layer1_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer1_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/'layer1_Norm_1' MathNormLayer output: [384,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer1_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (384, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/'layer1_weight_g' VariableLayer output: [384,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer1_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer1_weight_g', 'layer1_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/'layer1_truediv' CombineLayer output: [384,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer1_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer1_weight_v', 'layer1_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/'layer1_mul' CombineLayer output: [384,80,F|7]
*** root/.tmp_root:subnet/melgan:subnet/'layer3_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (384, 192, 10)}
*** root/.tmp_root:subnet/melgan:subnet/'layer3_weight' VariableLayer output: [384,192,F|10]
*** root/.tmp_root:subnet/melgan:subnet/'layer3_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer3_weight'}
*** root/.tmp_root:subnet/melgan:subnet/'layer3_Norm' MathNormLayer output: [384,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer3_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (384, 192, 10)}
*** root/.tmp_root:subnet/melgan:subnet/'layer3_weight_v' VariableLayer output: [384,192,F|10]
*** root/.tmp_root:subnet/melgan:subnet/'layer3_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer3_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/'layer3_Norm_1' MathNormLayer output: [384,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer3_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (384, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/'layer3_weight_g' VariableLayer output: [384,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer3_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer3_weight_g', 'layer3_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/'layer3_truediv' CombineLayer output: [384,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer3_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer3_weight_v', 'layer3_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/'layer3_mul' CombineLayer output: [384,192,F|10]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_weight' VariableLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_Norm' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_weight_v' VariableLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_Norm_1' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_weight_g' VariableLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_truediv' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_mul' CombineLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_weight' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_Norm' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_weight_v' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_Norm_1' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_weight_g' VariableLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_truediv' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_mul' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_weight' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_Norm' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_weight_v' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_Norm_1' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_weight_g' VariableLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_truediv' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_mul' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_weight' VariableLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_Norm' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_weight_v' VariableLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_Norm_1' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_weight_g' VariableLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_truediv' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_mul' CombineLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_weight' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_Norm' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_weight_v' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_Norm_1' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_weight_g' VariableLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_truediv' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_mul' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_weight' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_Norm' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_weight_v' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_Norm_1' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_weight_g' VariableLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_truediv' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_mul' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_weight' VariableLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_Norm' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_weight_v' VariableLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_Norm_1' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_weight_g' VariableLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_truediv' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_mul' CombineLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_weight' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_Norm' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_weight_v' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_Norm_1' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_weight_g' VariableLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_truediv' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_mul' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_weight' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_Norm' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_weight_v' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_Norm_1' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_weight_g' VariableLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_truediv' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_mul' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_weight' VariableLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_Norm' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_weight_v' VariableLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_Norm_1' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_weight_g' VariableLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_truediv' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_mul' CombineLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_weight' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_Norm' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_weight_v' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_Norm_1' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_weight_g' VariableLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_truediv' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_mul' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_weight' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_Norm' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 192, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_weight_v' VariableLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_Norm_1' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_weight_g' VariableLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_truediv' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_mul' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer9_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 96, 10)}
*** root/.tmp_root:subnet/melgan:subnet/'layer9_weight' VariableLayer output: [192,96,F|10]
*** root/.tmp_root:subnet/melgan:subnet/'layer9_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer9_weight'}
*** root/.tmp_root:subnet/melgan:subnet/'layer9_Norm' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer9_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 96, 10)}
*** root/.tmp_root:subnet/melgan:subnet/'layer9_weight_v' VariableLayer output: [192,96,F|10]
*** root/.tmp_root:subnet/melgan:subnet/'layer9_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer9_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/'layer9_Norm_1' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer9_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (192, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/'layer9_weight_g' VariableLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer9_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer9_weight_g', 'layer9_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/'layer9_truediv' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer9_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer9_weight_v', 'layer9_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/'layer9_mul' CombineLayer output: [192,96,F|10]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_weight' VariableLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_Norm' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_weight_v' VariableLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_Norm_1' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_weight_g' VariableLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_truediv' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_mul' CombineLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_weight' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_Norm' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_weight_v' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_Norm_1' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_weight_g' VariableLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_truediv' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_mul' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_weight' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_Norm' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_weight_v' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_Norm_1' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_weight_g' VariableLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_truediv' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_mul' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_weight' VariableLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_Norm' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_weight_v' VariableLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_Norm_1' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_weight_g' VariableLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_truediv' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_mul' CombineLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_weight' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_Norm' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_weight_v' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_Norm_1' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_weight_g' VariableLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_truediv' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_mul' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_weight' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_Norm' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_weight_v' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_Norm_1' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_weight_g' VariableLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_truediv' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_mul' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_weight' VariableLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_Norm' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_weight_v' VariableLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_Norm_1' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_weight_g' VariableLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_truediv' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_mul' CombineLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_weight' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_Norm' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_weight_v' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_Norm_1' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_weight_g' VariableLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_truediv' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_mul' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_weight' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_Norm' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_weight_v' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_Norm_1' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_weight_g' VariableLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_truediv' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_mul' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_weight' VariableLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_Norm' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_weight_v' VariableLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_Norm_1' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_weight_g' VariableLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_truediv' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_mul' CombineLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_weight' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_Norm' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_weight_v' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_Norm_1' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_weight_g' VariableLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_truediv' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_mul' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_weight' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_Norm' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 96, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_weight_v' VariableLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_Norm_1' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_weight_g' VariableLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_truediv' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_mul' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer15_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 48, 4)}
*** root/.tmp_root:subnet/melgan:subnet/'layer15_weight' VariableLayer output: [96,48,F|4]
*** root/.tmp_root:subnet/melgan:subnet/'layer15_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer15_weight'}
*** root/.tmp_root:subnet/melgan:subnet/'layer15_Norm' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer15_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 48, 4)}
*** root/.tmp_root:subnet/melgan:subnet/'layer15_weight_v' VariableLayer output: [96,48,F|4]
*** root/.tmp_root:subnet/melgan:subnet/'layer15_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer15_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/'layer15_Norm_1' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer15_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (96, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/'layer15_weight_g' VariableLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer15_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer15_weight_g', 'layer15_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/'layer15_truediv' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer15_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer15_weight_v', 'layer15_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/'layer15_mul' CombineLayer output: [96,48,F|4]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_weight' VariableLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_Norm' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_weight_v' VariableLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_Norm_1' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_weight_g' VariableLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_truediv' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_mul' CombineLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_weight' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_Norm' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_weight_v' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_Norm_1' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_weight_g' VariableLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_truediv' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_mul' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_weight' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_Norm' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_weight_v' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_Norm_1' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_weight_g' VariableLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_truediv' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_mul' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_weight' VariableLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_Norm' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_weight_v' VariableLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_Norm_1' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_weight_g' VariableLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_truediv' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_mul' CombineLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_weight' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_Norm' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_weight_v' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_Norm_1' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_weight_g' VariableLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_truediv' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_mul' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_weight' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_Norm' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_weight_v' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_Norm_1' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_weight_g' VariableLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_truediv' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_mul' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_weight' VariableLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_Norm' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_weight_v' VariableLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_Norm_1' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_weight_g' VariableLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_truediv' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_mul' CombineLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_weight' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_Norm' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_weight_v' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_Norm_1' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_weight_g' VariableLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_truediv' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_mul' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_weight' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_Norm' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_weight_v' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_Norm_1' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_weight_g' VariableLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_truediv' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_mul' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_weight' VariableLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_Norm' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 3)}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_weight_v' VariableLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_Norm_1' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_weight_g' VariableLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_truediv' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_mul' CombineLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_weight' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_Norm' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_weight_v' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_Norm_1' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_weight_g' VariableLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_truediv' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_mul' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_weight' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight'}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_Norm' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 48, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_weight_v' VariableLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_Norm_1' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (48, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_weight_g' VariableLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_truediv' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_mul' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer22_weight' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (4, 48, 7)}
*** root/.tmp_root:subnet/melgan:subnet/'layer22_weight' VariableLayer output: [4,48,F|7]
*** root/.tmp_root:subnet/melgan:subnet/'layer22_Norm' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer22_weight'}
*** root/.tmp_root:subnet/melgan:subnet/'layer22_Norm' MathNormLayer output: [4,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer22_weight_v' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (4, 48, 7)}
*** root/.tmp_root:subnet/melgan:subnet/'layer22_weight_v' VariableLayer output: [4,48,F|7]
*** root/.tmp_root:subnet/melgan:subnet/'layer22_Norm_1' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer22_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/'layer22_Norm_1' MathNormLayer output: [4,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer22_weight_g' layer dict: {'class': 'variable', 'add_batch_axis': False, 'shape': (4, 1, 1)}
*** root/.tmp_root:subnet/melgan:subnet/'layer22_weight_g' VariableLayer output: [4,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer22_truediv' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer22_weight_g', 'layer22_Norm_1']}
*** root/.tmp_root:subnet/melgan:subnet/'layer22_truediv' CombineLayer output: [4,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer22_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer22_weight_v', 'layer22_truediv']}
*** root/.tmp_root:subnet/melgan:subnet/'layer22_mul' CombineLayer output: [4,48,F|7]
*** root/.tmp_root:subnet/melgan:subnet/'layer1_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer1_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/'layer1_Norm_2' MathNormLayer output: [384,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer1_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer1_weight_g', 'layer1_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/'layer1_truediv_1' CombineLayer output: [384,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer1_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer1_weight_v', 'layer1_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/'layer1_mul_1' CombineLayer output: [384,80,F|7]
*** root/.tmp_root:subnet/melgan:subnet/'layer3_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer3_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/'layer3_Norm_2' MathNormLayer output: [384,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer3_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer3_weight_g', 'layer3_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/'layer3_truediv_1' CombineLayer output: [384,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer3_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer3_weight_v', 'layer3_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/'layer3_mul_1' CombineLayer output: [384,192,F|10]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_Norm_2' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_truediv_1' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer2_mul_1' CombineLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_Norm_2' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_truediv_1' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/stack:subnet/'layer4_mul_1' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_Norm_2' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_truediv_1' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer4:subnet/'skip_layer_mul_1' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_Norm_2' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_truediv_1' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer2_mul_1' CombineLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_Norm_2' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_truediv_1' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/stack:subnet/'layer4_mul_1' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_Norm_2' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_truediv_1' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer5:subnet/'skip_layer_mul_1' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_Norm_2' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_truediv_1' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer2_mul_1' CombineLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_Norm_2' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_truediv_1' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/stack:subnet/'layer4_mul_1' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_Norm_2' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_truediv_1' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer6:subnet/'skip_layer_mul_1' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_Norm_2' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_truediv_1' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer2_mul_1' CombineLayer output: [192,192,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_Norm_2' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_truediv_1' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/stack:subnet/'layer4_mul_1' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_Norm_2' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_truediv_1' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer7:subnet/'skip_layer_mul_1' CombineLayer output: [192,192,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer9_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer9_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/'layer9_Norm_2' MathNormLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer9_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer9_weight_g', 'layer9_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/'layer9_truediv_1' CombineLayer output: [192,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer9_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer9_weight_v', 'layer9_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/'layer9_mul_1' CombineLayer output: [192,96,F|10]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_Norm_2' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_truediv_1' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer2_mul_1' CombineLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_Norm_2' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_truediv_1' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/stack:subnet/'layer4_mul_1' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_Norm_2' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_truediv_1' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer10:subnet/'skip_layer_mul_1' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_Norm_2' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_truediv_1' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer2_mul_1' CombineLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_Norm_2' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_truediv_1' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/stack:subnet/'layer4_mul_1' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_Norm_2' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_truediv_1' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer11:subnet/'skip_layer_mul_1' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_Norm_2' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_truediv_1' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer2_mul_1' CombineLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_Norm_2' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_truediv_1' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/stack:subnet/'layer4_mul_1' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_Norm_2' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_truediv_1' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer12:subnet/'skip_layer_mul_1' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_Norm_2' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_truediv_1' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer2_mul_1' CombineLayer output: [96,96,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_Norm_2' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_truediv_1' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/stack:subnet/'layer4_mul_1' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_Norm_2' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_truediv_1' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer13:subnet/'skip_layer_mul_1' CombineLayer output: [96,96,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer15_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer15_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/'layer15_Norm_2' MathNormLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer15_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer15_weight_g', 'layer15_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/'layer15_truediv_1' CombineLayer output: [96,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer15_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer15_weight_v', 'layer15_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/'layer15_mul_1' CombineLayer output: [96,48,F|4]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_Norm_2' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_truediv_1' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer2_mul_1' CombineLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_Norm_2' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_truediv_1' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/stack:subnet/'layer4_mul_1' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_Norm_2' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_truediv_1' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer16:subnet/'skip_layer_mul_1' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_Norm_2' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_truediv_1' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer2_mul_1' CombineLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_Norm_2' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_truediv_1' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/stack:subnet/'layer4_mul_1' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_Norm_2' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_truediv_1' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer17:subnet/'skip_layer_mul_1' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_Norm_2' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_truediv_1' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer2_mul_1' CombineLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_Norm_2' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_truediv_1' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/stack:subnet/'layer4_mul_1' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_Norm_2' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_truediv_1' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer18:subnet/'skip_layer_mul_1' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer2_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_Norm_2' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer2_weight_g', 'layer2_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_truediv_1' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer2_weight_v', 'layer2_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer2_mul_1' CombineLayer output: [48,48,F|3]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer4_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_Norm_2' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer4_weight_g', 'layer4_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_truediv_1' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer4_weight_v', 'layer4_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/stack:subnet/'layer4_mul_1' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'skip_layer_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_Norm_2' MathNormLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['skip_layer_weight_g', 'skip_layer_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_truediv_1' CombineLayer output: [48,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['skip_layer_weight_v', 'skip_layer_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/layer19:subnet/'skip_layer_mul_1' CombineLayer output: [48,48,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer22_Norm_2' layer dict: {'class': 'math_norm', 'p': 2, 'keep_dims': True, 'axes': ['static:1', 'F'], 'from': 'layer22_weight_v'}
*** root/.tmp_root:subnet/melgan:subnet/'layer22_Norm_2' MathNormLayer output: [4,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer22_truediv_1' layer dict: {'class': 'combine', 'kind': 'truediv', 'from': ['layer22_weight_g', 'layer22_Norm_2']}
*** root/.tmp_root:subnet/melgan:subnet/'layer22_truediv_1' CombineLayer output: [4,1,F|1]
*** root/.tmp_root:subnet/melgan:subnet/'layer22_mul_1' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['layer22_weight_v', 'layer22_truediv_1']}
*** root/.tmp_root:subnet/melgan:subnet/'layer22_mul_1' CombineLayer output: [4,48,F|7]
*** root/.tmp_root:subnet/Cast_PQMF:subnet/'Cast_PQMF_unnamed_const' layer dict: {'class': 'constant', 'value': array([[ 6.84837762e-06, -8.61604274e-20, -1.57372949e-05,
1.11185474e-05, 1.61987245e-04, 4.59894990e-04,
7.72398151e-04, 8.38639994e-04, 4.98372346e-04,
-1.10215669e-18, 7.11952089e-05, 1.49111563e-03,
4.30004690e-03, 7.27269441e-03, 8.37646608e-03,
6.31989583e-03, 2.24644065e-03, 1.26586195e-19,
3.99743896e-03, 1.55457166e-02, 3.02382909e-02,
3.92091472e-02, 3.47761063e-02, 1.74983894e-02,
-5.26998738e-04, 2.48157347e-18, 3.44539554e-02,
1.02453039e-01, 1.83332388e-01, 2.43451169e-01,
2.52647189e-01, 2.00818326e-01, 1.04649892e-01,
1.49070848e-17, -7.59387617e-02, -1.02453039e-01,
-8.31792064e-02, -4.05271704e-02, -1.27228750e-03,
1.74983894e-02, 1.44047349e-02, 7.20260350e-18,
-1.25251102e-02, -1.55457166e-02, -9.65067136e-03,
-6.89103149e-04, 5.42338747e-03, 6.31989583e-03,
3.46964585e-03, 2.22662048e-18, -1.78113774e-03,
-1.49111563e-03, -1.71880439e-04, 9.22816437e-04,
1.20317728e-03, 8.38639994e-04, 3.19937790e-04,
-6.19814487e-19, -6.70973139e-05, -1.11185474e-05,
3.79931908e-05, 3.90753211e-05, 1.65334461e-05],
[ 1.65334461e-05, -1.24443201e-19, -3.79931908e-05,
1.11185474e-05, -6.70973139e-05, -4.59894990e-04,
-3.19937790e-04, 8.38639994e-04, 1.20317728e-03,
6.78645723e-19, 1.71880439e-04, 1.49111563e-03,
-1.78113774e-03, -7.27269441e-03, -3.46964585e-03,
6.31989583e-03, 5.42338747e-03, -7.59943711e-19,
9.65067136e-03, 1.55457166e-02, -1.25251102e-02,
-3.92091472e-02, -1.44047349e-02, 1.74983894e-02,
-1.27228750e-03, 1.24078674e-17, 8.31792064e-02,
1.02453039e-01, -7.59387617e-02, -2.43451169e-01,
-1.04649892e-01, 2.00818326e-01, 2.52647189e-01,
1.49070848e-17, -1.83332388e-01, -1.02453039e-01,
3.44539554e-02, 4.05271704e-02, 5.26998738e-04,
1.74983894e-02, 3.47761063e-02, 8.64555117e-17,
-3.02382909e-02, -1.55457166e-02, 3.99743896e-03,
6.89103149e-04, -2.24644065e-03, 6.31989583e-03,
8.37646608e-03, 1.87081137e-17, -4.30004690e-03,
-1.49111563e-03, 7.11952089e-05, -9.22816437e-04,
-4.98372346e-04, 8.38639994e-04, 7.72398151e-04,
1.35198609e-18, -1.61987245e-04, -1.11185474e-05,
-1.57372949e-05, -3.90753211e-05, -6.84837762e-06],
[-1.65334461e-05, 1.91171999e-20, 3.79931908e-05,
1.11185474e-05, 6.70973139e-05, -4.59894990e-04,
3.19937790e-04, 8.38639994e-04, -1.20317728e-03,
-3.39208620e-18, -1.71880439e-04, 1.49111563e-03,
1.78113774e-03, -7.27269441e-03, 3.46964585e-03,
6.31989583e-03, -5.42338747e-03, -1.94141486e-18,
-9.65067136e-03, 1.55457166e-02, 1.25251102e-02,
-3.92091472e-02, 1.44047349e-02, 1.74983894e-02,
1.27228750e-03, -8.93617306e-17, -8.31792064e-02,
1.02453039e-01, 7.59387617e-02, -2.43451169e-01,
1.04649892e-01, 2.00818326e-01, -2.52647189e-01,
-4.47212543e-17, 1.83332388e-01, -1.02453039e-01,
-3.44539554e-02, 4.05271704e-02, -5.26998738e-04,
1.74983894e-02, -3.47761063e-02, 3.84381550e-17,
3.02382909e-02, -1.55457166e-02, -3.99743896e-03,
6.89103149e-04, 2.24644065e-03, 6.31989583e-03,
-8.37646608e-03, -2.31613547e-17, 4.30004690e-03,
-1.49111563e-03, -7.11952089e-05, -9.22816437e-04,
4.98372346e-04, 8.38639994e-04, -7.72398151e-04,
-3.38010756e-18, 1.61987245e-04, -1.11185474e-05,
1.57372949e-05, -3.90753211e-05, 6.84837762e-06],
[-6.84837762e-06, 5.73999735e-20, 1.57372949e-05,
1.11185474e-05, -1.61987245e-04, 4.59894990e-04,
-7.72398151e-04, 8.38639994e-04, -4.98372346e-04,
3.84299347e-18, -7.11952089e-05, 1.49111563e-03,
-4.30004690e-03, 7.27269441e-03, -8.37646608e-03,
6.31989583e-03, -2.24644065e-03, -4.72716424e-18,
-3.99743896e-03, 1.55457166e-02, -3.02382909e-02,
3.92091472e-02, -3.47761063e-02, 1.74983894e-02,
5.26998738e-04, 4.46934081e-17, -3.44539554e-02,
1.02453039e-01, -1.83332388e-01, 2.43451169e-01,
-2.52647189e-01, 2.00818326e-01, -1.04649892e-01,
-4.47212543e-17, 7.59387617e-02, -1.02453039e-01,
8.31792064e-02, -4.05271704e-02, 1.27228750e-03,
1.74983894e-02, -1.44047349e-02, -1.10464190e-16,
1.25251102e-02, -1.55457166e-02, 9.65067136e-03,
-6.89103149e-04, -5.42338747e-03, 6.31989583e-03,
-3.46964585e-03, 2.49516544e-17, 1.78113774e-03,
-1.49111563e-03, 1.71880439e-04, 9.22816437e-04,
-1.20317728e-03, 8.38639994e-04, -3.19937790e-04,
-4.50282476e-19, 6.70973139e-05, -1.11185474e-05,
-3.79931908e-05, 3.90753211e-05, -1.65334461e-05]])}
*** root/.tmp_root:subnet/Cast_PQMF:subnet/'Cast_PQMF_unnamed_const' ConstantLayer output: [4,F|63]
*** root/.tmp_root:subnet/Cast_PQMF:subnet/'Cast' layer dict: {'class': 'cast', 'from': 'Cast_PQMF_unnamed_const', 'dtype': 'float32'}
*** root/.tmp_root:subnet/Cast_PQMF:subnet/'Cast' CastLayer output: [4,F|63]
*** root/.tmp_root:subnet/Cast_PQMF:subnet/'Cast_PQMF_unnamed_const_1' layer dict: {'class': 'constant', 'value': array([[ 1.65334461e-05, 3.90753211e-05, 3.79931908e-05,
-1.11185474e-05, -6.70973139e-05, -6.19814487e-19,
3.19937790e-04, 8.38639994e-04, 1.20317728e-03,
9.22816437e-04, -1.71880439e-04, -1.49111563e-03,
-1.78113774e-03, 2.22662048e-18, 3.46964585e-03,
6.31989583e-03, 5.42338747e-03, -6.89103149e-04,
-9.65067136e-03, -1.55457166e-02, -1.25251102e-02,
7.20260350e-18, 1.44047349e-02, 1.74983894e-02,
-1.27228750e-03, -4.05271704e-02, -8.31792064e-02,
-1.02453039e-01, -7.59387617e-02, 1.49070848e-17,
1.04649892e-01, 2.00818326e-01, 2.52647189e-01,
2.43451169e-01, 1.83332388e-01, 1.02453039e-01,
3.44539554e-02, 2.48157347e-18, -5.26998738e-04,
1.74983894e-02, 3.47761063e-02, 3.92091472e-02,
3.02382909e-02, 1.55457166e-02, 3.99743896e-03,
1.26586195e-19, 2.24644065e-03, 6.31989583e-03,
8.37646608e-03, 7.27269441e-03, 4.30004690e-03,
1.49111563e-03, 7.11952089e-05, -1.10215669e-18,
4.98372346e-04, 8.38639994e-04, 7.72398151e-04,
4.59894990e-04, 1.61987245e-04, 1.11185474e-05,
-1.57372949e-05, -8.61604274e-20, 6.84837762e-06],
[-6.84837762e-06, -3.90753211e-05, -1.57372949e-05,
-1.11185474e-05, -1.61987245e-04, 1.35198609e-18,
7.72398151e-04, 8.38639994e-04, -4.98372346e-04,
-9.22816437e-04, 7.11952089e-05, -1.49111563e-03,
-4.30004690e-03, 1.87081137e-17, 8.37646608e-03,
6.31989583e-03, -2.24644065e-03, 6.89103149e-04,
3.99743896e-03, -1.55457166e-02, -3.02382909e-02,
8.64555117e-17, 3.47761063e-02, 1.74983894e-02,
5.26998738e-04, 4.05271704e-02, 3.44539554e-02,
-1.02453039e-01, -1.83332388e-01, 1.49070848e-17,
2.52647189e-01, 2.00818326e-01, -1.04649892e-01,
-2.43451169e-01, -7.59387617e-02, 1.02453039e-01,
8.31792064e-02, 1.24078674e-17, -1.27228750e-03,
1.74983894e-02, -1.44047349e-02, -3.92091472e-02,
-1.25251102e-02, 1.55457166e-02, 9.65067136e-03,
-7.59943711e-19, 5.42338747e-03, 6.31989583e-03,
-3.46964585e-03, -7.27269441e-03, -1.78113774e-03,
1.49111563e-03, 1.71880439e-04, 6.78645723e-19,
1.20317728e-03, 8.38639994e-04, -3.19937790e-04,
-4.59894990e-04, -6.70973139e-05, 1.11185474e-05,
-3.79931908e-05, -1.24443201e-19, 1.65334461e-05],
[ 6.84837762e-06, -3.90753211e-05, 1.57372949e-05,
-1.11185474e-05, 1.61987245e-04, -3.38010756e-18,
-7.72398151e-04, 8.38639994e-04, 4.98372346e-04,
-9.22816437e-04, -7.11952089e-05, -1.49111563e-03,
4.30004690e-03, -2.31613547e-17, -8.37646608e-03,
6.31989583e-03, 2.24644065e-03, 6.89103149e-04,
-3.99743896e-03, -1.55457166e-02, 3.02382909e-02,
3.84381550e-17, -3.47761063e-02, 1.74983894e-02,
-5.26998738e-04, 4.05271704e-02, -3.44539554e-02,
-1.02453039e-01, 1.83332388e-01, -4.47212543e-17,
-2.52647189e-01, 2.00818326e-01, 1.04649892e-01,
-2.43451169e-01, 7.59387617e-02, 1.02453039e-01,
-8.31792064e-02, -8.93617306e-17, 1.27228750e-03,
1.74983894e-02, 1.44047349e-02, -3.92091472e-02,
1.25251102e-02, 1.55457166e-02, -9.65067136e-03,
-1.94141486e-18, -5.42338747e-03, 6.31989583e-03,
3.46964585e-03, -7.27269441e-03, 1.78113774e-03,
1.49111563e-03, -1.71880439e-04, -3.39208620e-18,
-1.20317728e-03, 8.38639994e-04, 3.19937790e-04,
-4.59894990e-04, 6.70973139e-05, 1.11185474e-05,
3.79931908e-05, 1.91171999e-20, -1.65334461e-05],
[-1.65334461e-05, 3.90753211e-05, -3.79931908e-05,
-1.11185474e-05, 6.70973139e-05, -4.50282476e-19,
-3.19937790e-04, 8.38639994e-04, -1.20317728e-03,
9.22816437e-04, 1.71880439e-04, -1.49111563e-03,
1.78113774e-03, 2.49516544e-17, -3.46964585e-03,
6.31989583e-03, -5.42338747e-03, -6.89103149e-04,
9.65067136e-03, -1.55457166e-02, 1.25251102e-02,
-1.10464190e-16, -1.44047349e-02, 1.74983894e-02,
1.27228750e-03, -4.05271704e-02, 8.31792064e-02,
-1.02453039e-01, 7.59387617e-02, -4.47212543e-17,
-1.04649892e-01, 2.00818326e-01, -2.52647189e-01,
2.43451169e-01, -1.83332388e-01, 1.02453039e-01,
-3.44539554e-02, 4.46934081e-17, 5.26998738e-04,
1.74983894e-02, -3.47761063e-02, 3.92091472e-02,
-3.02382909e-02, 1.55457166e-02, -3.99743896e-03,
-4.72716424e-18, -2.24644065e-03, 6.31989583e-03,
-8.37646608e-03, 7.27269441e-03, -4.30004690e-03,
1.49111563e-03, -7.11952089e-05, 3.84299347e-18,
-4.98372346e-04, 8.38639994e-04, -7.72398151e-04,
4.59894990e-04, -1.61987245e-04, 1.11185474e-05,
1.57372949e-05, 5.73999735e-20, -6.84837762e-06]])}
*** root/.tmp_root:subnet/Cast_PQMF:subnet/'Cast_PQMF_unnamed_const_1' ConstantLayer output: [4,F|63]
*** root/.tmp_root:subnet/Cast_PQMF:subnet/'Cast_1' layer dict: {'class': 'cast', 'from': 'Cast_PQMF_unnamed_const_1', 'dtype': 'float32'}
*** root/.tmp_root:subnet/Cast_PQMF:subnet/'Cast_1' CastLayer output: [4,F|63]
*** root/melgan:subnet/'layer0' layer dict: {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (3, 3), 'from': 'data'}
*** root/melgan:subnet/'layer0' PadLayer output: [B,F|80,T|'spatial:1:melgan/layer0']
*** root/melgan:subnet/'layer0' PadLayer check RETURNN outputs given Torch inputs/outputs ...
**** torch tensor func detach
*** root/melgan:subnet/'layer1' layer dict: {'class': 'conv', 'from': 'layer0', 'activation': None, 'with_bias': True, 'n_out': 384, 'filter_size': (7,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/'layer1' ConvLayer output: [B,T|'time:var:extern_data:data',F|384]
*** root/melgan:subnet/'layer1' ConvLayer importing params ['bias', 'weight'] ...)
WARNING:tensorflow:From /Users/az/Programmierung/import-parallel-wavegan/pytorch_to_returnn/torch/nn/modules/conv.py:103: Variable.load (from tensorflow.python.ops.variables) is deprecated and will be removed in a future version.
Instructions for updating:
Prefer Variable.assign which has equivalent behavior in 2.X.
*** root/melgan:subnet/'layer1' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/'layer2' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer1'}
*** root/melgan:subnet/'layer2' EvalLayer output: [B,T|'time:var:extern_data:data',F|384]
*** root/melgan:subnet/'layer2' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/'layer3' layer dict: {'class': 'transposed_conv', 'from': 'layer2', 'activation': None, 'with_bias': True, 'n_out': 192, 'filter_size': (10,), 'strides': (5,), 'padding': 'valid', 'output_padding': (1,), 'remove_padding': (3,)}
*** root/melgan:subnet/'layer3' TransposedConvLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/'layer3' TransposedConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/'layer3' TransposedConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer0' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'}
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer0' EvalLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer0' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer1' layer dict: {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (1, 1), 'from': 'layer0'}
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer1' PadLayer output: [B,T|'spatial:0:melgan/layer4/stack/layer1',F|192]
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer1' PadLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer2' layer dict: {'class': 'conv', 'from': 'layer1', 'activation': None, 'with_bias': True, 'n_out': 192, 'filter_size': (3,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer2' ConvLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer2' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer2' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer3' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'}
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer3' EvalLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer3' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer4' layer dict: {'class': 'conv', 'from': 'layer3', 'activation': None, 'with_bias': True, 'n_out': 192, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer4' ConvLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer4' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer4:subnet/stack:subnet/'layer4' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer4:subnet/stack:subnet/'output' layer dict: {'class': 'copy', 'from': 'layer4'}
*** root/melgan:subnet/layer4:subnet/stack:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer4:subnet/'stack' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer4:subnet/'stack' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer4:subnet/'skip_layer' layer dict: {'class': 'conv', 'from': 'data', 'activation': None, 'with_bias': True, 'n_out': 192, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer4:subnet/'skip_layer' ConvLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer4:subnet/'skip_layer' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer4:subnet/'skip_layer' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer4:subnet/'add' layer dict: {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']}
*** root/melgan:subnet/layer4:subnet/'add' CombineLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer4:subnet/'output' layer dict: {'class': 'copy', 'from': 'add'}
*** root/melgan:subnet/layer4:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/'layer4' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/'layer4' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer0' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'}
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer0' EvalLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer0' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer1' layer dict: {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (3, 3), 'from': 'layer0'}
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer1' PadLayer output: [B,T|'spatial:0:melgan/layer5/stack/layer1',F|192]
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer1' PadLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer2' layer dict: {'class': 'conv', 'from': 'layer1', 'activation': None, 'with_bias': True, 'n_out': 192, 'filter_size': (3,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (3,)}
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer2' ConvLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer2' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer2' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer3' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'}
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer3' EvalLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer3' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer4' layer dict: {'class': 'conv', 'from': 'layer3', 'activation': None, 'with_bias': True, 'n_out': 192, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer4' ConvLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer4' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer5:subnet/stack:subnet/'layer4' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer5:subnet/stack:subnet/'output' layer dict: {'class': 'copy', 'from': 'layer4'}
*** root/melgan:subnet/layer5:subnet/stack:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer5:subnet/'stack' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer5:subnet/'stack' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer5:subnet/'skip_layer' layer dict: {'class': 'conv', 'from': 'data', 'activation': None, 'with_bias': True, 'n_out': 192, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer5:subnet/'skip_layer' ConvLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer5:subnet/'skip_layer' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer5:subnet/'skip_layer' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer5:subnet/'add' layer dict: {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']}
*** root/melgan:subnet/layer5:subnet/'add' CombineLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer5:subnet/'output' layer dict: {'class': 'copy', 'from': 'add'}
*** root/melgan:subnet/layer5:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/'layer5' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/'layer5' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer0' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'}
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer0' EvalLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer0' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer1' layer dict: {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (9, 9), 'from': 'layer0'}
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer1' PadLayer output: [B,T|'spatial:0:melgan/layer6/stack/layer1',F|192]
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer1' PadLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer2' layer dict: {'class': 'conv', 'from': 'layer1', 'activation': None, 'with_bias': True, 'n_out': 192, 'filter_size': (3,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (9,)}
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer2' ConvLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer2' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer2' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer3' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'}
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer3' EvalLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer3' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer4' layer dict: {'class': 'conv', 'from': 'layer3', 'activation': None, 'with_bias': True, 'n_out': 192, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer4' ConvLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer4' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer6:subnet/stack:subnet/'layer4' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer6:subnet/stack:subnet/'output' layer dict: {'class': 'copy', 'from': 'layer4'}
*** root/melgan:subnet/layer6:subnet/stack:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer6:subnet/'stack' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer6:subnet/'stack' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer6:subnet/'skip_layer' layer dict: {'class': 'conv', 'from': 'data', 'activation': None, 'with_bias': True, 'n_out': 192, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer6:subnet/'skip_layer' ConvLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer6:subnet/'skip_layer' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer6:subnet/'skip_layer' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer6:subnet/'add' layer dict: {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']}
*** root/melgan:subnet/layer6:subnet/'add' CombineLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer6:subnet/'output' layer dict: {'class': 'copy', 'from': 'add'}
*** root/melgan:subnet/layer6:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/'layer6' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/'layer6' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer0' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'}
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer0' EvalLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer0' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer1' layer dict: {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (27, 27), 'from': 'layer0'}
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer1' PadLayer output: [B,T|'spatial:0:melgan/layer7/stack/layer1',F|192]
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer1' PadLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer2' layer dict: {'class': 'conv', 'from': 'layer1', 'activation': None, 'with_bias': True, 'n_out': 192, 'filter_size': (3,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (27,)}
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer2' ConvLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer2' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer2' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer3' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'}
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer3' EvalLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer3' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer4' layer dict: {'class': 'conv', 'from': 'layer3', 'activation': None, 'with_bias': True, 'n_out': 192, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer4' ConvLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer4' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer7:subnet/stack:subnet/'layer4' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer7:subnet/stack:subnet/'output' layer dict: {'class': 'copy', 'from': 'layer4'}
*** root/melgan:subnet/layer7:subnet/stack:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer7:subnet/'stack' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer7:subnet/'stack' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer7:subnet/'skip_layer' layer dict: {'class': 'conv', 'from': 'data', 'activation': None, 'with_bias': True, 'n_out': 192, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer7:subnet/'skip_layer' ConvLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer7:subnet/'skip_layer' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer7:subnet/'skip_layer' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer7:subnet/'add' layer dict: {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']}
*** root/melgan:subnet/layer7:subnet/'add' CombineLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/layer7:subnet/'output' layer dict: {'class': 'copy', 'from': 'add'}
*** root/melgan:subnet/layer7:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/'layer7' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/'layer7' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/'layer8' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer7'}
*** root/melgan:subnet/'layer8' EvalLayer output: [B,T|'spatial:0:melgan/layer3',F|192]
*** root/melgan:subnet/'layer8' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/'layer9' layer dict: {'class': 'transposed_conv', 'from': 'layer8', 'activation': None, 'with_bias': True, 'n_out': 96, 'filter_size': (10,), 'strides': (5,), 'padding': 'valid', 'output_padding': (1,), 'remove_padding': (3,)}
*** root/melgan:subnet/'layer9' TransposedConvLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/'layer9' TransposedConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/'layer9' TransposedConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer0' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'}
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer0' EvalLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer0' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer1' layer dict: {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (1, 1), 'from': 'layer0'}
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer1' PadLayer output: [B,T|'spatial:0:melgan/layer10/stack/layer1',F|96]
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer1' PadLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer2' layer dict: {'class': 'conv', 'from': 'layer1', 'activation': None, 'with_bias': True, 'n_out': 96, 'filter_size': (3,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer2' ConvLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer2' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer2' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer3' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'}
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer3' EvalLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer3' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer4' layer dict: {'class': 'conv', 'from': 'layer3', 'activation': None, 'with_bias': True, 'n_out': 96, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer4' ConvLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer4' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer10:subnet/stack:subnet/'layer4' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer10:subnet/stack:subnet/'output' layer dict: {'class': 'copy', 'from': 'layer4'}
*** root/melgan:subnet/layer10:subnet/stack:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer10:subnet/'stack' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer10:subnet/'stack' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer10:subnet/'skip_layer' layer dict: {'class': 'conv', 'from': 'data', 'activation': None, 'with_bias': True, 'n_out': 96, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer10:subnet/'skip_layer' ConvLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer10:subnet/'skip_layer' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer10:subnet/'skip_layer' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer10:subnet/'add' layer dict: {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']}
*** root/melgan:subnet/layer10:subnet/'add' CombineLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer10:subnet/'output' layer dict: {'class': 'copy', 'from': 'add'}
*** root/melgan:subnet/layer10:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/'layer10' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/'layer10' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer0' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'}
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer0' EvalLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer0' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer1' layer dict: {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (3, 3), 'from': 'layer0'}
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer1' PadLayer output: [B,T|'spatial:0:melgan/layer11/stack/layer1',F|96]
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer1' PadLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer2' layer dict: {'class': 'conv', 'from': 'layer1', 'activation': None, 'with_bias': True, 'n_out': 96, 'filter_size': (3,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (3,)}
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer2' ConvLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer2' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer2' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer3' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'}
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer3' EvalLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer3' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer4' layer dict: {'class': 'conv', 'from': 'layer3', 'activation': None, 'with_bias': True, 'n_out': 96, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer4' ConvLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer4' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer11:subnet/stack:subnet/'layer4' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer11:subnet/stack:subnet/'output' layer dict: {'class': 'copy', 'from': 'layer4'}
*** root/melgan:subnet/layer11:subnet/stack:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer11:subnet/'stack' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer11:subnet/'stack' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer11:subnet/'skip_layer' layer dict: {'class': 'conv', 'from': 'data', 'activation': None, 'with_bias': True, 'n_out': 96, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer11:subnet/'skip_layer' ConvLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer11:subnet/'skip_layer' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer11:subnet/'skip_layer' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer11:subnet/'add' layer dict: {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']}
*** root/melgan:subnet/layer11:subnet/'add' CombineLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer11:subnet/'output' layer dict: {'class': 'copy', 'from': 'add'}
*** root/melgan:subnet/layer11:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/'layer11' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/'layer11' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer0' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'}
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer0' EvalLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer0' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer1' layer dict: {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (9, 9), 'from': 'layer0'}
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer1' PadLayer output: [B,T|'spatial:0:melgan/layer12/stack/layer1',F|96]
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer1' PadLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer2' layer dict: {'class': 'conv', 'from': 'layer1', 'activation': None, 'with_bias': True, 'n_out': 96, 'filter_size': (3,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (9,)}
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer2' ConvLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer2' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer2' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer3' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'}
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer3' EvalLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer3' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer4' layer dict: {'class': 'conv', 'from': 'layer3', 'activation': None, 'with_bias': True, 'n_out': 96, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer4' ConvLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer4' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer12:subnet/stack:subnet/'layer4' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer12:subnet/stack:subnet/'output' layer dict: {'class': 'copy', 'from': 'layer4'}
*** root/melgan:subnet/layer12:subnet/stack:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer12:subnet/'stack' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer12:subnet/'stack' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer12:subnet/'skip_layer' layer dict: {'class': 'conv', 'from': 'data', 'activation': None, 'with_bias': True, 'n_out': 96, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer12:subnet/'skip_layer' ConvLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer12:subnet/'skip_layer' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer12:subnet/'skip_layer' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer12:subnet/'add' layer dict: {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']}
*** root/melgan:subnet/layer12:subnet/'add' CombineLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer12:subnet/'output' layer dict: {'class': 'copy', 'from': 'add'}
*** root/melgan:subnet/layer12:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/'layer12' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/'layer12' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer0' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'}
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer0' EvalLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer0' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer1' layer dict: {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (27, 27), 'from': 'layer0'}
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer1' PadLayer output: [B,T|'spatial:0:melgan/layer13/stack/layer1',F|96]
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer1' PadLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer2' layer dict: {'class': 'conv', 'from': 'layer1', 'activation': None, 'with_bias': True, 'n_out': 96, 'filter_size': (3,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (27,)}
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer2' ConvLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer2' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer2' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer3' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'}
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer3' EvalLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer3' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer4' layer dict: {'class': 'conv', 'from': 'layer3', 'activation': None, 'with_bias': True, 'n_out': 96, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer4' ConvLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer4' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer13:subnet/stack:subnet/'layer4' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer13:subnet/stack:subnet/'output' layer dict: {'class': 'copy', 'from': 'layer4'}
*** root/melgan:subnet/layer13:subnet/stack:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer13:subnet/'stack' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer13:subnet/'stack' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer13:subnet/'skip_layer' layer dict: {'class': 'conv', 'from': 'data', 'activation': None, 'with_bias': True, 'n_out': 96, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer13:subnet/'skip_layer' ConvLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer13:subnet/'skip_layer' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer13:subnet/'skip_layer' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer13:subnet/'add' layer dict: {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']}
*** root/melgan:subnet/layer13:subnet/'add' CombineLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/layer13:subnet/'output' layer dict: {'class': 'copy', 'from': 'add'}
*** root/melgan:subnet/layer13:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/'layer13' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/'layer13' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/'layer14' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer13'}
*** root/melgan:subnet/'layer14' EvalLayer output: [B,T|'spatial:0:melgan/layer9',F|96]
*** root/melgan:subnet/'layer14' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/'layer15' layer dict: {'class': 'transposed_conv', 'from': 'layer14', 'activation': None, 'with_bias': True, 'n_out': 48, 'filter_size': (4,), 'strides': (2,), 'padding': 'valid', 'output_padding': (0,), 'remove_padding': (1,)}
*** root/melgan:subnet/'layer15' TransposedConvLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/'layer15' TransposedConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/'layer15' TransposedConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer0' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'}
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer0' EvalLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer0' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer1' layer dict: {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (1, 1), 'from': 'layer0'}
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer1' PadLayer output: [B,T|'spatial:0:melgan/layer16/stack/layer1',F|48]
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer1' PadLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer2' layer dict: {'class': 'conv', 'from': 'layer1', 'activation': None, 'with_bias': True, 'n_out': 48, 'filter_size': (3,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer2' ConvLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer2' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer2' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer3' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'}
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer3' EvalLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer3' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer4' layer dict: {'class': 'conv', 'from': 'layer3', 'activation': None, 'with_bias': True, 'n_out': 48, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer4' ConvLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer4' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer16:subnet/stack:subnet/'layer4' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer16:subnet/stack:subnet/'output' layer dict: {'class': 'copy', 'from': 'layer4'}
*** root/melgan:subnet/layer16:subnet/stack:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer16:subnet/'stack' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer16:subnet/'stack' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer16:subnet/'skip_layer' layer dict: {'class': 'conv', 'from': 'data', 'activation': None, 'with_bias': True, 'n_out': 48, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer16:subnet/'skip_layer' ConvLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer16:subnet/'skip_layer' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer16:subnet/'skip_layer' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer16:subnet/'add' layer dict: {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']}
*** root/melgan:subnet/layer16:subnet/'add' CombineLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer16:subnet/'output' layer dict: {'class': 'copy', 'from': 'add'}
*** root/melgan:subnet/layer16:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/'layer16' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/'layer16' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer0' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'}
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer0' EvalLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer0' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer1' layer dict: {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (3, 3), 'from': 'layer0'}
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer1' PadLayer output: [B,T|'spatial:0:melgan/layer17/stack/layer1',F|48]
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer1' PadLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer2' layer dict: {'class': 'conv', 'from': 'layer1', 'activation': None, 'with_bias': True, 'n_out': 48, 'filter_size': (3,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (3,)}
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer2' ConvLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer2' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer2' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer3' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'}
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer3' EvalLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer3' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer4' layer dict: {'class': 'conv', 'from': 'layer3', 'activation': None, 'with_bias': True, 'n_out': 48, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer4' ConvLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer4' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer17:subnet/stack:subnet/'layer4' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer17:subnet/stack:subnet/'output' layer dict: {'class': 'copy', 'from': 'layer4'}
*** root/melgan:subnet/layer17:subnet/stack:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer17:subnet/'stack' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer17:subnet/'stack' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer17:subnet/'skip_layer' layer dict: {'class': 'conv', 'from': 'data', 'activation': None, 'with_bias': True, 'n_out': 48, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer17:subnet/'skip_layer' ConvLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer17:subnet/'skip_layer' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer17:subnet/'skip_layer' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer17:subnet/'add' layer dict: {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']}
*** root/melgan:subnet/layer17:subnet/'add' CombineLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer17:subnet/'output' layer dict: {'class': 'copy', 'from': 'add'}
*** root/melgan:subnet/layer17:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/'layer17' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/'layer17' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer0' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'}
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer0' EvalLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer0' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer1' layer dict: {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (9, 9), 'from': 'layer0'}
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer1' PadLayer output: [B,T|'spatial:0:melgan/layer18/stack/layer1',F|48]
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer1' PadLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer2' layer dict: {'class': 'conv', 'from': 'layer1', 'activation': None, 'with_bias': True, 'n_out': 48, 'filter_size': (3,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (9,)}
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer2' ConvLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer2' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer2' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer3' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'}
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer3' EvalLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer3' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer4' layer dict: {'class': 'conv', 'from': 'layer3', 'activation': None, 'with_bias': True, 'n_out': 48, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer4' ConvLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer4' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer18:subnet/stack:subnet/'layer4' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer18:subnet/stack:subnet/'output' layer dict: {'class': 'copy', 'from': 'layer4'}
*** root/melgan:subnet/layer18:subnet/stack:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer18:subnet/'stack' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer18:subnet/'stack' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer18:subnet/'skip_layer' layer dict: {'class': 'conv', 'from': 'data', 'activation': None, 'with_bias': True, 'n_out': 48, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer18:subnet/'skip_layer' ConvLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer18:subnet/'skip_layer' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer18:subnet/'skip_layer' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer18:subnet/'add' layer dict: {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']}
*** root/melgan:subnet/layer18:subnet/'add' CombineLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer18:subnet/'output' layer dict: {'class': 'copy', 'from': 'add'}
*** root/melgan:subnet/layer18:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/'layer18' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/'layer18' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer0' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'}
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer0' EvalLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer0' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer1' layer dict: {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (27, 27), 'from': 'layer0'}
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer1' PadLayer output: [B,T|'spatial:0:melgan/layer19/stack/layer1',F|48]
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer1' PadLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer2' layer dict: {'class': 'conv', 'from': 'layer1', 'activation': None, 'with_bias': True, 'n_out': 48, 'filter_size': (3,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (27,)}
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer2' ConvLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer2' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer2' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer3' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'}
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer3' EvalLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer3' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer4' layer dict: {'class': 'conv', 'from': 'layer3', 'activation': None, 'with_bias': True, 'n_out': 48, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer4' ConvLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer4' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer19:subnet/stack:subnet/'layer4' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer19:subnet/stack:subnet/'output' layer dict: {'class': 'copy', 'from': 'layer4'}
*** root/melgan:subnet/layer19:subnet/stack:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer19:subnet/'stack' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer19:subnet/'stack' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer19:subnet/'skip_layer' layer dict: {'class': 'conv', 'from': 'data', 'activation': None, 'with_bias': True, 'n_out': 48, 'filter_size': (1,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/layer19:subnet/'skip_layer' ConvLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer19:subnet/'skip_layer' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/layer19:subnet/'skip_layer' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/layer19:subnet/'add' layer dict: {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']}
*** root/melgan:subnet/layer19:subnet/'add' CombineLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/layer19:subnet/'output' layer dict: {'class': 'copy', 'from': 'add'}
*** root/melgan:subnet/layer19:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/'layer19' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/'layer19' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/'layer20' layer dict: {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer19'}
*** root/melgan:subnet/'layer20' EvalLayer output: [B,T|'spatial:0:melgan/layer15',F|48]
*** root/melgan:subnet/'layer20' EvalLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/'layer21' layer dict: {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (3, 3), 'from': 'layer20'}
*** root/melgan:subnet/'layer21' PadLayer output: [B,T|'spatial:0:melgan/layer21',F|48]
*** root/melgan:subnet/'layer21' PadLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/'layer22' layer dict: {'class': 'conv', 'from': 'layer21', 'activation': None, 'with_bias': True, 'n_out': 4, 'filter_size': (7,), 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/melgan:subnet/'layer22' ConvLayer output: [B,T|'spatial:0:melgan/layer15',F|4]
*** root/melgan:subnet/'layer22' ConvLayer importing params ['bias', 'weight'] ...)
*** root/melgan:subnet/'layer22' ConvLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/'layer23' layer dict: {'class': 'activation', 'activation': 'tanh', 'from': 'layer22'}
*** root/melgan:subnet/'layer23' ActivationLayer output: [B,T|'spatial:0:melgan/layer15',F|4]
*** root/melgan:subnet/'layer23' ActivationLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/melgan:subnet/'output' layer dict: {'class': 'copy', 'from': 'layer23'}
*** root/melgan:subnet/'output' CopyLayer output: [B,T|'spatial:0:melgan/layer15',F|4]
*** root/'melgan' SubnetworkLayer output: [B,T|'spatial:0:melgan/layer15',F|4]
*** root/'melgan' SubnetworkLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/'PQMF_Cast_unnamed_const' layer dict: {'class': 'constant', 'value': array(4, dtype=int32)}
*** root/'PQMF_Cast_unnamed_const' ConstantLayer output: []
*** root/'PQMF_Cast' layer dict: {'class': 'cast', 'from': 'PQMF_Cast_unnamed_const', 'dtype': 'float32'}
*** root/'PQMF_Cast' CastLayer output: []
*** root/'PQMF_updown_filter' layer dict: {'class': 'constant', 'value': array([[[1., 0., 0., 0.],
[0., 0., 0., 0.],
[0., 0., 0., 0.],
[0., 0., 0., 0.]],
[[0., 0., 0., 0.],
[1., 0., 0., 0.],
[0., 0., 0., 0.],
[0., 0., 0., 0.]],
[[0., 0., 0., 0.],
[0., 0., 0., 0.],
[1., 0., 0., 0.],
[0., 0., 0., 0.]],
[[0., 0., 0., 0.],
[0., 0., 0., 0.],
[0., 0., 0., 0.],
[1., 0., 0., 0.]]], dtype=float32)}
*** root/'PQMF_updown_filter' ConstantLayer output: [4,4,F|4]
*** root/'PQMF_mul' layer dict: {'class': 'combine', 'kind': 'mul', 'from': ['PQMF_updown_filter', 'PQMF_Cast']}
*** root/'PQMF_mul' CombineLayer output: [4,4,F|4]
*** root/'PQMF_FunctionalConvTransposed1d' layer dict: {'class': 'transposed_conv', 'from': 'melgan', 'n_out': 4, 'activation': None, 'with_bias': False, 'bias': None, 'filter_size': (4,), 'filter': 'PQMF_mul', 'filter_perm': {'static:0': 'F', 'static:1': 'static:1', 'F': 'static:0'}, 'padding': 'valid', 'output_padding': (0,), 'remove_padding': (0,), 'strides': (4,)}
*** root/'PQMF_FunctionalConvTransposed1d' TransposedConvLayer output: [B,T|'spatial:0:PQMF_FunctionalConvTransposed1d',F|4]
*** root/'pad_fn' layer dict: {'class': 'pad', 'mode': 'constant', 'axes': 'spatial', 'padding': (31, 31), 'from': 'PQMF_FunctionalConvTransposed1d', 'value': 0.0}
*** root/'pad_fn' PadLayer output: [B,T|'spatial:0:pad_fn',F|4]
*** root/'pad_fn' PadLayer check RETURNN outputs given Torch inputs/outputs ...
*** root/'PQMF_synthesis_filter' layer dict: {'class': 'constant', 'value': array([[[ 1.65334459e-05, 3.90753194e-05, 3.79931917e-05,
-1.11185473e-05, -6.70973168e-05, -6.19814498e-19,
3.19937797e-04, 8.38639971e-04, 1.20317726e-03,
9.22816456e-04, -1.71880441e-04, -1.49111566e-03,
-1.78113778e-03, 2.22662049e-18, 3.46964574e-03,
6.31989585e-03, 5.42338751e-03, -6.89103152e-04,
-9.65067092e-03, -1.55457165e-02, -1.25251105e-02,
7.20260322e-18, 1.44047346e-02, 1.74983889e-02,
-1.27228745e-03, -4.05271687e-02, -8.31792057e-02,
-1.02453038e-01, -7.59387612e-02, 1.49070850e-17,
1.04649894e-01, 2.00818330e-01, 2.52647191e-01,
2.43451163e-01, 1.83332384e-01, 1.02453038e-01,
3.44539545e-02, 2.48157348e-18, -5.26998716e-04,
1.74983889e-02, 3.47761065e-02, 3.92091461e-02,
3.02382912e-02, 1.55457165e-02, 3.99743905e-03,
1.26586189e-19, 2.24644062e-03, 6.31989585e-03,
8.37646611e-03, 7.27269426e-03, 4.30004671e-03,
1.49111566e-03, 7.11952089e-05, -1.10215664e-18,
4.98372363e-04, 8.38639971e-04, 7.72398140e-04,
4.59895004e-04, 1.61987249e-04, 1.11185473e-05,
-1.57372942e-05, -8.61604251e-20, 6.84837778e-06],
[-6.84837778e-06, -3.90753194e-05, -1.57372942e-05,
-1.11185473e-05, -1.61987249e-04, 1.35198608e-18,
7.72398140e-04, 8.38639971e-04, -4.98372363e-04,
-9.22816456e-04, 7.11952089e-05, -1.49111566e-03,
-4.30004671e-03, 1.87081140e-17, 8.37646611e-03,
6.31989585e-03, -2.24644062e-03, 6.89103152e-04,
3.99743905e-03, -1.55457165e-02, -3.02382912e-02,
8.64555147e-17, 3.47761065e-02, 1.74983889e-02,
5.26998716e-04, 4.05271687e-02, 3.44539545e-02,
-1.02453038e-01, -1.83332384e-01, 1.49070850e-17,
2.52647191e-01, 2.00818330e-01, -1.04649894e-01,
-2.43451163e-01, -7.59387612e-02, 1.02453038e-01,
8.31792057e-02, 1.24078672e-17, -1.27228745e-03,
1.74983889e-02, -1.44047346e-02, -3.92091461e-02,
-1.25251105e-02, 1.55457165e-02, 9.65067092e-03,
-7.59943702e-19, 5.42338751e-03, 6.31989585e-03,
-3.46964574e-03, -7.27269426e-03, -1.78113778e-03,
1.49111566e-03, 1.71880441e-04, 6.78645703e-19,
1.20317726e-03, 8.38639971e-04, -3.19937797e-04,
-4.59895004e-04, -6.70973168e-05, 1.11185473e-05,
-3.79931917e-05, -1.24443197e-19, 1.65334459e-05],
[ 6.84837778e-06, -3.90753194e-05, 1.57372942e-05,
-1.11185473e-05, 1.61987249e-04, -3.38010752e-18,
-7.72398140e-04, 8.38639971e-04, 4.98372363e-04,
-9.22816456e-04, -7.11952089e-05, -1.49111566e-03,
4.30004671e-03, -2.31613549e-17, -8.37646611e-03,
6.31989585e-03, 2.24644062e-03, 6.89103152e-04,
-3.99743905e-03, -1.55457165e-02, 3.02382912e-02,
3.84381566e-17, -3.47761065e-02, 1.74983889e-02,
-5.26998716e-04, 4.05271687e-02, -3.44539545e-02,
-1.02453038e-01, 1.83332384e-01, -4.47212551e-17,
-2.52647191e-01, 2.00818330e-01, 1.04649894e-01,
-2.43451163e-01, 7.59387612e-02, 1.02453038e-01,
-8.31792057e-02, -8.93617311e-17, 1.27228745e-03,
1.74983889e-02, 1.44047346e-02, -3.92091461e-02,
1.25251105e-02, 1.55457165e-02, -9.65067092e-03,
-1.94141482e-18, -5.42338751e-03, 6.31989585e-03,
3.46964574e-03, -7.27269426e-03, 1.78113778e-03,
1.49111566e-03, -1.71880441e-04, -3.39208613e-18,
-1.20317726e-03, 8.38639971e-04, 3.19937797e-04,
-4.59895004e-04, 6.70973168e-05, 1.11185473e-05,
3.79931917e-05, 1.91172006e-20, -1.65334459e-05],
[-1.65334459e-05, 3.90753194e-05, -3.79931917e-05,
-1.11185473e-05, 6.70973168e-05, -4.50282487e-19,
-3.19937797e-04, 8.38639971e-04, -1.20317726e-03,
9.22816456e-04, 1.71880441e-04, -1.49111566e-03,
1.78113778e-03, 2.49516550e-17, -3.46964574e-03,
6.31989585e-03, -5.42338751e-03, -6.89103152e-04,
9.65067092e-03, -1.55457165e-02, 1.25251105e-02,
-1.10464187e-16, -1.44047346e-02, 1.74983889e-02,
1.27228745e-03, -4.05271687e-02, 8.31792057e-02,
-1.02453038e-01, 7.59387612e-02, -4.47212551e-17,
-1.04649894e-01, 2.00818330e-01, -2.52647191e-01,
2.43451163e-01, -1.83332384e-01, 1.02453038e-01,
-3.44539545e-02, 4.46934089e-17, 5.26998716e-04,
1.74983889e-02, -3.47761065e-02, 3.92091461e-02,
-3.02382912e-02, 1.55457165e-02, -3.99743905e-03,
-4.72716432e-18, -2.24644062e-03, 6.31989585e-03,
-8.37646611e-03, 7.27269426e-03, -4.30004671e-03,
1.49111566e-03, -7.11952089e-05, 3.84299344e-18,
-4.98372363e-04, 8.38639971e-04, -7.72398140e-04,
4.59895004e-04, -1.61987249e-04, 1.11185473e-05,
1.57372942e-05, 5.73999756e-20, -6.84837778e-06]]],
dtype=float32)}
*** root/'PQMF_synthesis_filter' ConstantLayer output: [1,4,F|63]
*** root/'PQMF_FunctionalConv1d' layer dict: {'class': 'conv', 'from': 'pad_fn', 'n_out': 1, 'activation': None, 'with_bias': False, 'bias': None, 'filter_size': (63,), 'filter': 'PQMF_synthesis_filter', 'filter_perm': {'static:0': 'F', 'static:1': 'static:1', 'F': 'static:0'}, 'padding': 'valid', 'strides': (1,), 'dilation_rate': (1,)}
*** root/'PQMF_FunctionalConv1d' ConvLayer output: [B,T|'spatial:0:PQMF_FunctionalConvTransposed1d',F|1]
*** root/'output' layer dict: {'class': 'copy', 'from': 'PQMF_FunctionalConv1d'}
*** root/'output' CopyLayer output: [B,T|'spatial:0:PQMF_FunctionalConvTransposed1d',F|1]
RETURNN output: Data(name='output_output', shape=(None, 1), batch_shape_meta=[B,T|'spatial:0:PQMF_FunctionalConvTransposed1d',F|1]) axis map RETURNN->Torch {0: 0, 2: 1, 1: 2}
>>>> Module naming hierarchy:
.tmp_root: (hidden)
data: None -> None
melgan: <ModuleEntry <Sequential>> -> <TensorEntry name:? tensor:(1,4,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|4] axes {0:0,2:1,1:2}>
data: None -> None
layer0: <ModuleEntry <ReflectionPad1d>> -> <TensorEntry name:? tensor:(1,80,80) returnn_data:'layer0_output' [B,F|80,T|'spatial:1:melgan/layer0'] axes id>
layer1: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,384,80) returnn_data:'layer1_output' [B,T|'time:var:extern_data:data',F|384] axes {0:0,2:1,1:2}>
layer2: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,384,80) returnn_data:'layer2_output' [B,T|'time:var:extern_data:data',F|384] axes {0:0,2:1,1:2}>
layer3: <ModuleEntry <ConvTranspose1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer3_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer4: <ModuleEntry <ResidualStack>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
data: None -> None
stack: <ModuleEntry <Sequential>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
data: None -> None
layer0: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer0_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer1: <ModuleEntry <ReflectionPad1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer1_output' [B,T|'spatial:0:melgan/layer4/stack/layer1',F|192] axes {0:0,2:1,1:2}>
layer2: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer2_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer3: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer3_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer4: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
skip_layer: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'skip_layer_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
add: <ModuleEntry <BinaryOperator>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer5: <ModuleEntry <ResidualStack>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
data: None -> None
stack: <ModuleEntry <Sequential>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
data: None -> None
layer0: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer0_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer1: <ModuleEntry <ReflectionPad1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer1_output' [B,T|'spatial:0:melgan/layer5/stack/layer1',F|192] axes {0:0,2:1,1:2}>
layer2: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer2_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer3: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer3_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer4: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
skip_layer: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'skip_layer_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
add: <ModuleEntry <BinaryOperator>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer6: <ModuleEntry <ResidualStack>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
data: None -> None
stack: <ModuleEntry <Sequential>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
data: None -> None
layer0: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer0_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer1: <ModuleEntry <ReflectionPad1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer1_output' [B,T|'spatial:0:melgan/layer6/stack/layer1',F|192] axes {0:0,2:1,1:2}>
layer2: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer2_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer3: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer3_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer4: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
skip_layer: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'skip_layer_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
add: <ModuleEntry <BinaryOperator>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer7: <ModuleEntry <ResidualStack>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
data: None -> None
stack: <ModuleEntry <Sequential>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
data: None -> None
layer0: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer0_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer1: <ModuleEntry <ReflectionPad1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer1_output' [B,T|'spatial:0:melgan/layer7/stack/layer1',F|192] axes {0:0,2:1,1:2}>
layer2: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer2_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer3: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer3_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer4: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
skip_layer: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'skip_layer_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
add: <ModuleEntry <BinaryOperator>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer8: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,192,80) returnn_data:'layer8_output' [B,T|'spatial:0:melgan/layer3',F|192] axes {0:0,2:1,1:2}>
layer9: <ModuleEntry <ConvTranspose1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer9_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer10: <ModuleEntry <ResidualStack>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
data: None -> None
stack: <ModuleEntry <Sequential>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
data: None -> None
layer0: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer0_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer1: <ModuleEntry <ReflectionPad1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer1_output' [B,T|'spatial:0:melgan/layer10/stack/layer1',F|96] axes {0:0,2:1,1:2}>
layer2: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer2_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer3: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer3_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer4: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
skip_layer: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'skip_layer_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
add: <ModuleEntry <BinaryOperator>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer11: <ModuleEntry <ResidualStack>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
data: None -> None
stack: <ModuleEntry <Sequential>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
data: None -> None
layer0: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer0_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer1: <ModuleEntry <ReflectionPad1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer1_output' [B,T|'spatial:0:melgan/layer11/stack/layer1',F|96] axes {0:0,2:1,1:2}>
layer2: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer2_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer3: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer3_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer4: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
skip_layer: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'skip_layer_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
add: <ModuleEntry <BinaryOperator>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer12: <ModuleEntry <ResidualStack>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
data: None -> None
stack: <ModuleEntry <Sequential>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
data: None -> None
layer0: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer0_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer1: <ModuleEntry <ReflectionPad1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer1_output' [B,T|'spatial:0:melgan/layer12/stack/layer1',F|96] axes {0:0,2:1,1:2}>
layer2: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer2_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer3: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer3_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer4: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
skip_layer: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'skip_layer_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
add: <ModuleEntry <BinaryOperator>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer13: <ModuleEntry <ResidualStack>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
data: None -> None
stack: <ModuleEntry <Sequential>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
data: None -> None
layer0: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer0_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer1: <ModuleEntry <ReflectionPad1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer1_output' [B,T|'spatial:0:melgan/layer13/stack/layer1',F|96] axes {0:0,2:1,1:2}>
layer2: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer2_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer3: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer3_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer4: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
skip_layer: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'skip_layer_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
add: <ModuleEntry <BinaryOperator>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer14: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,96,80) returnn_data:'layer14_output' [B,T|'spatial:0:melgan/layer9',F|96] axes {0:0,2:1,1:2}>
layer15: <ModuleEntry <ConvTranspose1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer15_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer16: <ModuleEntry <ResidualStack>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
data: None -> None
stack: <ModuleEntry <Sequential>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
data: None -> None
layer0: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer0_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer1: <ModuleEntry <ReflectionPad1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer1_output' [B,T|'spatial:0:melgan/layer16/stack/layer1',F|48] axes {0:0,2:1,1:2}>
layer2: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer2_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer3: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer3_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer4: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
skip_layer: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'skip_layer_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
add: <ModuleEntry <BinaryOperator>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer17: <ModuleEntry <ResidualStack>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
data: None -> None
stack: <ModuleEntry <Sequential>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
data: None -> None
layer0: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer0_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer1: <ModuleEntry <ReflectionPad1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer1_output' [B,T|'spatial:0:melgan/layer17/stack/layer1',F|48] axes {0:0,2:1,1:2}>
layer2: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer2_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer3: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer3_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer4: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
skip_layer: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'skip_layer_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
add: <ModuleEntry <BinaryOperator>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer18: <ModuleEntry <ResidualStack>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
data: None -> None
stack: <ModuleEntry <Sequential>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
data: None -> None
layer0: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer0_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer1: <ModuleEntry <ReflectionPad1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer1_output' [B,T|'spatial:0:melgan/layer18/stack/layer1',F|48] axes {0:0,2:1,1:2}>
layer2: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer2_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer3: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer3_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer4: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
skip_layer: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'skip_layer_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
add: <ModuleEntry <BinaryOperator>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer19: <ModuleEntry <ResidualStack>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
data: None -> None
stack: <ModuleEntry <Sequential>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
data: None -> None
layer0: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer0_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer1: <ModuleEntry <ReflectionPad1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer1_output' [B,T|'spatial:0:melgan/layer19/stack/layer1',F|48] axes {0:0,2:1,1:2}>
layer2: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer2_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer3: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer3_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer4: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
skip_layer: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'skip_layer_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
add: <ModuleEntry <BinaryOperator>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer20: <ModuleEntry <LeakyReLU>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer20_output' [B,T|'spatial:0:melgan/layer15',F|48] axes {0:0,2:1,1:2}>
layer21: <ModuleEntry <ReflectionPad1d>> -> <TensorEntry name:? tensor:(1,48,80) returnn_data:'layer21_output' [B,T|'spatial:0:melgan/layer21',F|48] axes {0:0,2:1,1:2}>
layer22: <ModuleEntry <Conv1d>> -> <TensorEntry name:? tensor:(1,4,80) returnn_data:'layer22_output' [B,T|'spatial:0:melgan/layer15',F|4] axes {0:0,2:1,1:2}>
layer23: <ModuleEntry <Tanh>> -> <TensorEntry name:? tensor:(1,4,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|4] axes {0:0,2:1,1:2}>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,4,80) returnn_data:'output_output' [B,T|'spatial:0:melgan/layer15',F|4] axes {0:0,2:1,1:2}>
PQMF_Cast: <ModuleEntry <Cast>> -> <TensorEntry name:? tensor:() returnn_data:'PQMF_Cast_output' [] axes id>
PQMF_Cast_unnamed_const: <ModuleEntry <Constant>> -> <TensorEntry name:'value' tensor:() returnn_data:'PQMF_Cast_unnamed_const_const' [] axes id>
PQMF_mul: <ModuleEntry <BinaryOperator>> -> <TensorEntry name:? tensor:(4,4,4) returnn_data:'PQMF_mul_output' [4,4,F|4] axes id>
PQMF_updown_filter: <ModuleEntry <Constant>> -> <TensorEntry name:'updown_filter' tensor:(4,4,4) returnn_data:'PQMF_updown_filter_const' [4,4,F|4] axes id>
PQMF_FunctionalConvTransposed1d: <ModuleEntry <FunctionalConvTransposed1d>> -> <TensorEntry name:? tensor:(1,4,80) returnn_data:'PQMF_FunctionalConvTransposed1d_output' [B,T|'spatial:0:PQMF_FunctionalConvTransposed1d',F|4] axes {0:0,2:1,1:2}>
pad_fn: <ModuleEntry <ConstantPad1d>> -> <TensorEntry name:? tensor:(1,4,80) returnn_data:'pad_fn_output' [B,T|'spatial:0:pad_fn',F|4] axes {0:0,2:1,1:2}>
PQMF_FunctionalConv1d: <ModuleEntry <FunctionalConv1d>> -> <TensorEntry name:? tensor:(1,1,80) returnn_data:'output_output' [B,T|'spatial:0:PQMF_FunctionalConvTransposed1d',F|1] axes {0:0,2:1,1:2}>
PQMF_synthesis_filter: <ModuleEntry <Constant>> -> <TensorEntry name:'synthesis_filter' tensor:(1,4,63) returnn_data:'PQMF_synthesis_filter_const' [1,4,F|63] axes id>
output: <ModuleEntry <Copy>> -> <TensorEntry name:? tensor:(1,1,80) returnn_data:'output_output' [B,T|'spatial:0:PQMF_FunctionalConvTransposed1d',F|1] axes {0:0,2:1,1:2}>
>>>> RETURNN net dict:
{
'melgan': {
'class': 'subnetwork',
'from': 'data',
'subnetwork': {
'layer0': {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (3, 3), 'from': 'data'},
'layer1': {
'class': 'conv',
'from': 'layer0',
'activation': None,
'with_bias': True,
'n_out': 384,
'filter_size': (7,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'layer2': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer1'},
'layer3': {
'class': 'transposed_conv',
'from': 'layer2',
'activation': None,
'with_bias': True,
'n_out': 192,
'filter_size': (10,),
'strides': (5,),
'padding': 'valid',
'output_padding': (1,),
'remove_padding': (3,)
},
'layer4': {
'class': 'subnetwork',
'from': 'layer3',
'subnetwork': {
'stack': {
'class': 'subnetwork',
'from': 'data',
'subnetwork': {
'layer0': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'},
'layer1': {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (1, 1), 'from': 'layer0'},
'layer2': {
'class': 'conv',
'from': 'layer1',
'activation': None,
'with_bias': True,
'n_out': 192,
'filter_size': (3,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'layer3': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'},
'layer4': {
'class': 'conv',
'from': 'layer3',
'activation': None,
'with_bias': True,
'n_out': 192,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'output': {'class': 'copy', 'from': 'layer4'}
}
},
'skip_layer': {
'class': 'conv',
'from': 'data',
'activation': None,
'with_bias': True,
'n_out': 192,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'add': {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']},
'output': {'class': 'copy', 'from': 'add'}
}
},
'layer5': {
'class': 'subnetwork',
'from': 'layer4',
'subnetwork': {
'stack': {
'class': 'subnetwork',
'from': 'data',
'subnetwork': {
'layer0': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'},
'layer1': {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (3, 3), 'from': 'layer0'},
'layer2': {
'class': 'conv',
'from': 'layer1',
'activation': None,
'with_bias': True,
'n_out': 192,
'filter_size': (3,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (3,)
},
'layer3': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'},
'layer4': {
'class': 'conv',
'from': 'layer3',
'activation': None,
'with_bias': True,
'n_out': 192,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'output': {'class': 'copy', 'from': 'layer4'}
}
},
'skip_layer': {
'class': 'conv',
'from': 'data',
'activation': None,
'with_bias': True,
'n_out': 192,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'add': {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']},
'output': {'class': 'copy', 'from': 'add'}
}
},
'layer6': {
'class': 'subnetwork',
'from': 'layer5',
'subnetwork': {
'stack': {
'class': 'subnetwork',
'from': 'data',
'subnetwork': {
'layer0': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'},
'layer1': {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (9, 9), 'from': 'layer0'},
'layer2': {
'class': 'conv',
'from': 'layer1',
'activation': None,
'with_bias': True,
'n_out': 192,
'filter_size': (3,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (9,)
},
'layer3': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'},
'layer4': {
'class': 'conv',
'from': 'layer3',
'activation': None,
'with_bias': True,
'n_out': 192,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'output': {'class': 'copy', 'from': 'layer4'}
}
},
'skip_layer': {
'class': 'conv',
'from': 'data',
'activation': None,
'with_bias': True,
'n_out': 192,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'add': {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']},
'output': {'class': 'copy', 'from': 'add'}
}
},
'layer7': {
'class': 'subnetwork',
'from': 'layer6',
'subnetwork': {
'stack': {
'class': 'subnetwork',
'from': 'data',
'subnetwork': {
'layer0': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'},
'layer1': {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (27, 27), 'from': 'layer0'},
'layer2': {
'class': 'conv',
'from': 'layer1',
'activation': None,
'with_bias': True,
'n_out': 192,
'filter_size': (3,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (27,)
},
'layer3': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'},
'layer4': {
'class': 'conv',
'from': 'layer3',
'activation': None,
'with_bias': True,
'n_out': 192,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'output': {'class': 'copy', 'from': 'layer4'}
}
},
'skip_layer': {
'class': 'conv',
'from': 'data',
'activation': None,
'with_bias': True,
'n_out': 192,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'add': {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']},
'output': {'class': 'copy', 'from': 'add'}
}
},
'layer8': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer7'},
'layer9': {
'class': 'transposed_conv',
'from': 'layer8',
'activation': None,
'with_bias': True,
'n_out': 96,
'filter_size': (10,),
'strides': (5,),
'padding': 'valid',
'output_padding': (1,),
'remove_padding': (3,)
},
'layer10': {
'class': 'subnetwork',
'from': 'layer9',
'subnetwork': {
'stack': {
'class': 'subnetwork',
'from': 'data',
'subnetwork': {
'layer0': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'},
'layer1': {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (1, 1), 'from': 'layer0'},
'layer2': {
'class': 'conv',
'from': 'layer1',
'activation': None,
'with_bias': True,
'n_out': 96,
'filter_size': (3,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'layer3': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'},
'layer4': {
'class': 'conv',
'from': 'layer3',
'activation': None,
'with_bias': True,
'n_out': 96,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'output': {'class': 'copy', 'from': 'layer4'}
}
},
'skip_layer': {
'class': 'conv',
'from': 'data',
'activation': None,
'with_bias': True,
'n_out': 96,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'add': {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']},
'output': {'class': 'copy', 'from': 'add'}
}
},
'layer11': {
'class': 'subnetwork',
'from': 'layer10',
'subnetwork': {
'stack': {
'class': 'subnetwork',
'from': 'data',
'subnetwork': {
'layer0': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'},
'layer1': {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (3, 3), 'from': 'layer0'},
'layer2': {
'class': 'conv',
'from': 'layer1',
'activation': None,
'with_bias': True,
'n_out': 96,
'filter_size': (3,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (3,)
},
'layer3': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'},
'layer4': {
'class': 'conv',
'from': 'layer3',
'activation': None,
'with_bias': True,
'n_out': 96,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'output': {'class': 'copy', 'from': 'layer4'}
}
},
'skip_layer': {
'class': 'conv',
'from': 'data',
'activation': None,
'with_bias': True,
'n_out': 96,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'add': {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']},
'output': {'class': 'copy', 'from': 'add'}
}
},
'layer12': {
'class': 'subnetwork',
'from': 'layer11',
'subnetwork': {
'stack': {
'class': 'subnetwork',
'from': 'data',
'subnetwork': {
'layer0': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'},
'layer1': {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (9, 9), 'from': 'layer0'},
'layer2': {
'class': 'conv',
'from': 'layer1',
'activation': None,
'with_bias': True,
'n_out': 96,
'filter_size': (3,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (9,)
},
'layer3': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'},
'layer4': {
'class': 'conv',
'from': 'layer3',
'activation': None,
'with_bias': True,
'n_out': 96,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'output': {'class': 'copy', 'from': 'layer4'}
}
},
'skip_layer': {
'class': 'conv',
'from': 'data',
'activation': None,
'with_bias': True,
'n_out': 96,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'add': {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']},
'output': {'class': 'copy', 'from': 'add'}
}
},
'layer13': {
'class': 'subnetwork',
'from': 'layer12',
'subnetwork': {
'stack': {
'class': 'subnetwork',
'from': 'data',
'subnetwork': {
'layer0': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'},
'layer1': {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (27, 27), 'from': 'layer0'},
'layer2': {
'class': 'conv',
'from': 'layer1',
'activation': None,
'with_bias': True,
'n_out': 96,
'filter_size': (3,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (27,)
},
'layer3': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'},
'layer4': {
'class': 'conv',
'from': 'layer3',
'activation': None,
'with_bias': True,
'n_out': 96,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'output': {'class': 'copy', 'from': 'layer4'}
}
},
'skip_layer': {
'class': 'conv',
'from': 'data',
'activation': None,
'with_bias': True,
'n_out': 96,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'add': {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']},
'output': {'class': 'copy', 'from': 'add'}
}
},
'layer14': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer13'},
'layer15': {
'class': 'transposed_conv',
'from': 'layer14',
'activation': None,
'with_bias': True,
'n_out': 48,
'filter_size': (4,),
'strides': (2,),
'padding': 'valid',
'output_padding': (0,),
'remove_padding': (1,)
},
'layer16': {
'class': 'subnetwork',
'from': 'layer15',
'subnetwork': {
'stack': {
'class': 'subnetwork',
'from': 'data',
'subnetwork': {
'layer0': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'},
'layer1': {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (1, 1), 'from': 'layer0'},
'layer2': {
'class': 'conv',
'from': 'layer1',
'activation': None,
'with_bias': True,
'n_out': 48,
'filter_size': (3,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'layer3': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'},
'layer4': {
'class': 'conv',
'from': 'layer3',
'activation': None,
'with_bias': True,
'n_out': 48,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'output': {'class': 'copy', 'from': 'layer4'}
}
},
'skip_layer': {
'class': 'conv',
'from': 'data',
'activation': None,
'with_bias': True,
'n_out': 48,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'add': {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']},
'output': {'class': 'copy', 'from': 'add'}
}
},
'layer17': {
'class': 'subnetwork',
'from': 'layer16',
'subnetwork': {
'stack': {
'class': 'subnetwork',
'from': 'data',
'subnetwork': {
'layer0': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'},
'layer1': {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (3, 3), 'from': 'layer0'},
'layer2': {
'class': 'conv',
'from': 'layer1',
'activation': None,
'with_bias': True,
'n_out': 48,
'filter_size': (3,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (3,)
},
'layer3': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'},
'layer4': {
'class': 'conv',
'from': 'layer3',
'activation': None,
'with_bias': True,
'n_out': 48,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'output': {'class': 'copy', 'from': 'layer4'}
}
},
'skip_layer': {
'class': 'conv',
'from': 'data',
'activation': None,
'with_bias': True,
'n_out': 48,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'add': {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']},
'output': {'class': 'copy', 'from': 'add'}
}
},
'layer18': {
'class': 'subnetwork',
'from': 'layer17',
'subnetwork': {
'stack': {
'class': 'subnetwork',
'from': 'data',
'subnetwork': {
'layer0': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'},
'layer1': {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (9, 9), 'from': 'layer0'},
'layer2': {
'class': 'conv',
'from': 'layer1',
'activation': None,
'with_bias': True,
'n_out': 48,
'filter_size': (3,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (9,)
},
'layer3': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'},
'layer4': {
'class': 'conv',
'from': 'layer3',
'activation': None,
'with_bias': True,
'n_out': 48,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'output': {'class': 'copy', 'from': 'layer4'}
}
},
'skip_layer': {
'class': 'conv',
'from': 'data',
'activation': None,
'with_bias': True,
'n_out': 48,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'add': {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']},
'output': {'class': 'copy', 'from': 'add'}
}
},
'layer19': {
'class': 'subnetwork',
'from': 'layer18',
'subnetwork': {
'stack': {
'class': 'subnetwork',
'from': 'data',
'subnetwork': {
'layer0': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'data'},
'layer1': {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (27, 27), 'from': 'layer0'},
'layer2': {
'class': 'conv',
'from': 'layer1',
'activation': None,
'with_bias': True,
'n_out': 48,
'filter_size': (3,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (27,)
},
'layer3': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer2'},
'layer4': {
'class': 'conv',
'from': 'layer3',
'activation': None,
'with_bias': True,
'n_out': 48,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'output': {'class': 'copy', 'from': 'layer4'}
}
},
'skip_layer': {
'class': 'conv',
'from': 'data',
'activation': None,
'with_bias': True,
'n_out': 48,
'filter_size': (1,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'add': {'class': 'combine', 'kind': 'add', 'from': ['stack', 'skip_layer']},
'output': {'class': 'copy', 'from': 'add'}
}
},
'layer20': {'class': 'eval', 'eval': 'tf.nn.leaky_relu(source(0), alpha=0.2)', 'from': 'layer19'},
'layer21': {'class': 'pad', 'mode': 'reflect', 'axes': 'spatial', 'padding': (3, 3), 'from': 'layer20'},
'layer22': {
'class': 'conv',
'from': 'layer21',
'activation': None,
'with_bias': True,
'n_out': 4,
'filter_size': (7,),
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'layer23': {'class': 'activation', 'activation': 'tanh', 'from': 'layer22'},
'output': {'class': 'copy', 'from': 'layer23'}
}
},
'PQMF_Cast': {'class': 'cast', 'from': 'PQMF_Cast_unnamed_const', 'dtype': 'float32'},
'PQMF_Cast_unnamed_const': {'class': 'constant', 'value': numpy.array(4, dtype=numpy.int32)},
'PQMF_mul': {'class': 'combine', 'kind': 'mul', 'from': ['PQMF_updown_filter', 'PQMF_Cast']},
'PQMF_updown_filter': {
'class': 'constant',
'value': numpy.array([
[[1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0]],
[[0.0, 0.0, 0.0, 0.0], [1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0]],
[[0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0], [1.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0]],
[[0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0], [1.0, 0.0, 0.0, 0.0]]
], dtype=numpy.float32)
},
'PQMF_FunctionalConvTransposed1d': {
'class': 'transposed_conv',
'from': 'melgan',
'n_out': 4,
'activation': None,
'with_bias': False,
'bias': None,
'filter_size': (4,),
'filter': 'PQMF_mul',
'filter_perm': {'static:0': 'F', 'static:1': 'static:1', 'F': 'static:0'},
'padding': 'valid',
'output_padding': (0,),
'remove_padding': (0,),
'strides': (4,)
},
'pad_fn': {
'class': 'pad',
'mode': 'constant',
'axes': 'spatial',
'padding': (31, 31),
'from': 'PQMF_FunctionalConvTransposed1d',
'value': 0.0
},
'PQMF_FunctionalConv1d': {
'class': 'conv',
'from': 'pad_fn',
'n_out': 1,
'activation': None,
'with_bias': False,
'bias': None,
'filter_size': (63,),
'filter': 'PQMF_synthesis_filter',
'filter_perm': {'static:0': 'F', 'static:1': 'static:1', 'F': 'static:0'},
'padding': 'valid',
'strides': (1,),
'dilation_rate': (1,)
},
'PQMF_synthesis_filter': {
'class': 'constant',
'value': numpy.array([
[
[
1.653344588703476e-05,
3.907531936420128e-05,
3.7993191654095426e-05,
-1.1118547263322398e-05,
-6.70973167871125e-05,
-6.19814497996956e-19,
0.000319937797030434,
0.0008386399713344872,
0.0012031772639602423,
0.0009228164562955499,
-0.00017188044148497283,
-0.0014911156613379717,
-0.0017811377765610814,
2.2266204893018784e-18,
0.00346964574418962,
0.006319895852357149,
0.005423387512564659,
-0.0006891031516715884,
-0.009650670923292637,
-0.015545716509222984,
-0.012525110505521297,
7.202603219092196e-18,
0.01440473459661007,
0.01749838888645172,
-0.0012722874525934458,
-0.04052716866135597,
-0.08317920565605164,
-0.10245303809642792,
-0.07593876123428345,
1.490708503996114e-17,
0.10464989393949509,
0.20081833004951477,
0.25264719128608704,
0.24345116317272186,
0.1833323836326599,
0.10245303809642792,
0.03445395454764366,
2.481573477317513e-18,
-0.0005269987159408629,
0.01749838888645172,
0.03477610647678375,
0.03920914605259895,
0.03023829124867916,
0.015545716509222984,
0.003997439052909613,
1.265861891747689e-19,
0.0022464406210929155,
0.006319895852357149,
0.008376466110348701,
0.007272694259881973,
0.0043000467121601105,
0.0014911156613379717,
7.119520887499675e-05,
-1.1021566446688507e-18,
0.0004983723629266024,
0.0008386399713344872,
0.0007723981398157775,
0.0004598950035870075,
0.0001619872491573915,
1.1118547263322398e-05,
-1.5737294233986177e-05,
-8.616042507323048e-20,
6.848377779533621e-06
],
[
-6.848377779533621e-06,
-3.907531936420128e-05,
-1.5737294233986177e-05,
-1.1118547263322398e-05,
-0.0001619872491573915,
1.351986075970332e-18,
0.0007723981398157775,
0.0008386399713344872,
-0.0004983723629266024,
-0.0009228164562955499,
7.119520887499675e-05,
-0.0014911156613379717,
-0.0043000467121601105,
1.8708113957901537e-17,
0.008376466110348701,
0.006319895852357149,
-0.0022464406210929155,
0.0006891031516715884,
0.003997439052909613,
-0.015545716509222984,
-0.03023829124867916,
8.645551472572356e-17,
0.03477610647678375,
0.01749838888645172,
0.0005269987159408629,
0.04052716866135597,
0.03445395454764366,
-0.10245303809642792,
-0.1833323836326599,
1.490708503996114e-17,
0.25264719128608704,
0.20081833004951477,
-0.10464989393949509,
-0.24345116317272186,
-0.07593876123428345,
0.10245303809642792,
0.08317920565605164,
1.2407867179792413e-17,
-0.0012722874525934458,
0.01749838888645172,
-0.01440473459661007,
-0.03920914605259895,
-0.012525110505521297,
0.015545716509222984,
0.009650670923292637,
-7.599437017507494e-19,
0.005423387512564659,
0.006319895852357149,
-0.00346964574418962,
-0.007272694259881973,
-0.0017811377765610814,
0.0014911156613379717,
0.00017188044148497283,
6.78645702812047e-19,
0.0012031772639602423,
0.0008386399713344872,
-0.000319937797030434,
-0.0004598950035870075,
-6.70973167871125e-05,
1.1118547263322398e-05,
-3.7993191654095426e-05,
-1.244431968521913e-19,
1.653344588703476e-05
],
[
6.848377779533621e-06,
-3.907531936420128e-05,
1.5737294233986177e-05,
-1.1118547263322398e-05,
0.0001619872491573915,
-3.3801075166899776e-18,
-0.0007723981398157775,
0.0008386399713344872,
0.0004983723629266024,
-0.0009228164562955499,
-7.119520887499675e-05,
-0.0014911156613379717,
0.0043000467121601105,
-2.3161354936505294e-17,
-0.008376466110348701,
0.006319895852357149,
0.0022464406210929155,
0.0006891031516715884,
-0.003997439052909613,
-0.015545716509222984,
0.03023829124867916,
3.8438156623053137e-17,
-0.03477610647678375,
0.01749838888645172,
-0.0005269987159408629,
0.04052716866135597,
-0.03445395454764366,
-0.10245303809642792,
0.1833323836326599,
-4.4721255119883424e-17,
-0.25264719128608704,
0.20081833004951477,
0.10464989393949509,
-0.24345116317272186,
0.07593876123428345,
0.10245303809642792,
-0.08317920565605164,
-8.936173108986737e-17,
0.0012722874525934458,
0.01749838888645172,
0.01440473459661007,
-0.03920914605259895,
0.012525110505521297,
0.015545716509222984,
-0.009650670923292637,
-1.9414148179481886e-18,
-0.005423387512564659,
0.006319895852357149,
0.00346964574418962,
-0.007272694259881973,
0.0017811377765610814,
0.0014911156613379717,
-0.00017188044148497283,
-3.392086125935511e-18,
-0.0012031772639602423,
0.0008386399713344872,
0.000319937797030434,
-0.0004598950035870075,
6.70973167871125e-05,
1.1118547263322398e-05,
3.7993191654095426e-05,
1.9117200550086035e-20,
-1.653344588703476e-05
],
[
-1.653344588703476e-05,
3.907531936420128e-05,
-3.7993191654095426e-05,
-1.1118547263322398e-05,
6.70973167871125e-05,
-4.502824872857176e-19,
-0.000319937797030434,
0.0008386399713344872,
-0.0012031772639602423,
0.0009228164562955499,
0.00017188044148497283,
-0.0014911156613379717,
0.0017811377765610814,
2.4951655023478314e-17,
-0.00346964574418962,
0.006319895852357149,
-0.005423387512564659,
-0.0006891031516715884,
0.009650670923292637,
-0.015545716509222984,
0.012525110505521297,
-1.1046418715961388e-16,
-0.01440473459661007,
0.01749838888645172,
0.0012722874525934458,
-0.04052716866135597,
0.08317920565605164,
-0.10245303809642792,
0.07593876123428345,
-4.4721255119883424e-17,
-0.10464989393949509,
0.20081833004951477,
-0.25264719128608704,
0.24345116317272186,
-0.1833323836326599,
0.10245303809642792,
-0.03445395454764366,
4.469340891174244e-17,
0.0005269987159408629,
0.01749838888645172,
-0.03477610647678375,
0.03920914605259895,
-0.03023829124867916,
0.015545716509222984,
-0.003997439052909613,
-4.7271643199925296e-18,
-0.0022464406210929155,
0.006319895852357149,
-0.008376466110348701,
0.007272694259881973,
-0.0043000467121601105,
0.0014911156613379717,
-7.119520887499675e-05,
3.842993444776436e-18,
-0.0004983723629266024,
0.0008386399713344872,
-0.0007723981398157775,
0.0004598950035870075,
-0.0001619872491573915,
1.1118547263322398e-05,
1.5737294233986177e-05,
5.739997556022112e-20,
-6.848377779533621e-06
]
]
], dtype=numpy.float32)
},
'output': {'class': 'copy', 'from': 'PQMF_FunctionalConv1d'}
}
>>>> Root module calls:
{
'melgan': <CallEntry #1 <ModuleEntry <Sequential>>>,
'PQMF_Cast': <CallEntry #0 <ModuleEntry <Cast>>>,
'PQMF_Cast_unnamed_const': <CallEntry #1 <ModuleEntry <Constant>>>,
'PQMF_mul': <CallEntry #0 <ModuleEntry <BinaryOperator>>>,
'PQMF_updown_filter': <CallEntry #1 <ModuleEntry <Constant>>>,
'PQMF_FunctionalConvTransposed1d': <CallEntry #0 <ModuleEntry <FunctionalConvTransposed1d>>>,
'pad_fn': <CallEntry #0 <ModuleEntry <ConstantPad1d>>>,
'PQMF_FunctionalConv1d': <CallEntry #0 <ModuleEntry <FunctionalConv1d>>>,
'PQMF_synthesis_filter': <CallEntry #1 <ModuleEntry <Constant>>>,
'output': <CallEntry #None <ModuleEntry <Copy>>>
}
>>>> Modules with params:
{
'melgan.layer1': <Conv1d>,
'melgan.layer3': <ConvTranspose1d>,
'melgan.layer4.stack.layer2': <Conv1d>,
'melgan.layer4.stack.layer4': <Conv1d>,
'melgan.layer4.skip_layer': <Conv1d>,
'melgan.layer5.stack.layer2': <Conv1d>,
'melgan.layer5.stack.layer4': <Conv1d>,
'melgan.layer5.skip_layer': <Conv1d>,
'melgan.layer6.stack.layer2': <Conv1d>,
'melgan.layer6.stack.layer4': <Conv1d>,
'melgan.layer6.skip_layer': <Conv1d>,
'melgan.layer7.stack.layer2': <Conv1d>,
'melgan.layer7.stack.layer4': <Conv1d>,
'melgan.layer7.skip_layer': <Conv1d>,
'melgan.layer9': <ConvTranspose1d>,
'melgan.layer10.stack.layer2': <Conv1d>,
'melgan.layer10.stack.layer4': <Conv1d>,
'melgan.layer10.skip_layer': <Conv1d>,
'melgan.layer11.stack.layer2': <Conv1d>,
'melgan.layer11.stack.layer4': <Conv1d>,
'melgan.layer11.skip_layer': <Conv1d>,
'melgan.layer12.stack.layer2': <Conv1d>,
'melgan.layer12.stack.layer4': <Conv1d>,
'melgan.layer12.skip_layer': <Conv1d>,
'melgan.layer13.stack.layer2': <Conv1d>,
'melgan.layer13.stack.layer4': <Conv1d>,
'melgan.layer13.skip_layer': <Conv1d>,
'melgan.layer15': <ConvTranspose1d>,
'melgan.layer16.stack.layer2': <Conv1d>,
'melgan.layer16.stack.layer4': <Conv1d>,
'melgan.layer16.skip_layer': <Conv1d>,
'melgan.layer17.stack.layer2': <Conv1d>,
'melgan.layer17.stack.layer4': <Conv1d>,
'melgan.layer17.skip_layer': <Conv1d>,
'melgan.layer18.stack.layer2': <Conv1d>,
'melgan.layer18.stack.layer4': <Conv1d>,
'melgan.layer18.skip_layer': <Conv1d>,
'melgan.layer19.stack.layer2': <Conv1d>,
'melgan.layer19.stack.layer4': <Conv1d>,
'melgan.layer19.skip_layer': <Conv1d>,
'melgan.layer22': <Conv1d>
}
Output shape: (1, 16000, 1)
Output seq lens: [16000]
Output shape (converted to Torch): (1, 1, 16000)
>>>> Looks good!
RETURNN network layer topology:
extern data: data: Data(shape=(80, None), time_dim_axis=2, feature_dim_axis=1, batch_shape_meta=[B,F|80,T|'time:var:extern_data:data'])
used data keys: ['data']
layers:
layer subnetwork '.tmp_root' #: 80
layer cast 'PQMF_Cast' #: unknown
layer constant 'PQMF_Cast_unnamed_const' #: unknown
layer conv 'PQMF_FunctionalConv1d' #: 1
layer transposed_conv 'PQMF_FunctionalConvTransposed1d' #: 4
layer combine 'PQMF_mul' #: 4
layer constant 'PQMF_synthesis_filter' #: 63
layer constant 'PQMF_updown_filter' #: 4
layer source 'data' #: 80
layer subnetwork 'melgan' #: 4
layer copy 'output' #: 1
layer pad 'pad_fn' #: 4
net params #: 2128852
net trainable params: [<tf.Variable 'melgan/layer1/W:0' shape=(7, 80, 384) dtype=float32>, <tf.Variable 'melgan/layer1/bias:0' shape=(384,) dtype=float32>, <tf.Variable 'melgan/layer10/skip_layer/W:0' shape=(1, 96, 96) dtype=float32>, <tf.Variable 'melgan/layer10/skip_layer/bias:0' shape=(96,) dtype=float32>, <tf.Variable 'melgan/layer10/stack/layer2/W:0' shape=(3, 96, 96) dtype=float32>, <tf.Variable 'melgan/layer10/stack/layer2/bias:0' shape=(96,) dtype=float32>, <tf.Variable 'melgan/layer10/stack/layer4/W:0' shape=(1, 96, 96) dtype=float32>, <tf.Variable 'melgan/layer10/stack/layer4/bias:0' shape=(96,) dtype=float32>, <tf.Variable 'melgan/layer11/skip_layer/W:0' shape=(1, 96, 96) dtype=float32>, <tf.Variable 'melgan/layer11/skip_layer/bias:0' shape=(96,) dtype=float32>, <tf.Variable 'melgan/layer11/stack/layer2/W:0' shape=(3, 96, 96) dtype=float32>, <tf.Variable 'melgan/layer11/stack/layer2/bias:0' shape=(96,) dtype=float32>, <tf.Variable 'melgan/layer11/stack/layer4/W:0' shape=(1, 96, 96) dtype=float32>, <tf.Variable 'melgan/layer11/stack/layer4/bias:0' shape=(96,) dtype=float32>, <tf.Variable 'melgan/layer12/skip_layer/W:0' shape=(1, 96, 96) dtype=float32>, <tf.Variable 'melgan/layer12/skip_layer/bias:0' shape=(96,) dtype=float32>, <tf.Variable 'melgan/layer12/stack/layer2/W:0' shape=(3, 96, 96) dtype=float32>, <tf.Variable 'melgan/layer12/stack/layer2/bias:0' shape=(96,) dtype=float32>, <tf.Variable 'melgan/layer12/stack/layer4/W:0' shape=(1, 96, 96) dtype=float32>, <tf.Variable 'melgan/layer12/stack/layer4/bias:0' shape=(96,) dtype=float32>, <tf.Variable 'melgan/layer13/skip_layer/W:0' shape=(1, 96, 96) dtype=float32>, <tf.Variable 'melgan/layer13/skip_layer/bias:0' shape=(96,) dtype=float32>, <tf.Variable 'melgan/layer13/stack/layer2/W:0' shape=(3, 96, 96) dtype=float32>, <tf.Variable 'melgan/layer13/stack/layer2/bias:0' shape=(96,) dtype=float32>, <tf.Variable 'melgan/layer13/stack/layer4/W:0' shape=(1, 96, 96) dtype=float32>, <tf.Variable 'melgan/layer13/stack/layer4/bias:0' shape=(96,) dtype=float32>, <tf.Variable 'melgan/layer15/W_native_transposed_conv:0' shape=(4, 1, 48, 96) dtype=float32>, <tf.Variable 'melgan/layer15/bias:0' shape=(48,) dtype=float32>, <tf.Variable 'melgan/layer16/skip_layer/W:0' shape=(1, 48, 48) dtype=float32>, <tf.Variable 'melgan/layer16/skip_layer/bias:0' shape=(48,) dtype=float32>, <tf.Variable 'melgan/layer16/stack/layer2/W:0' shape=(3, 48, 48) dtype=float32>, <tf.Variable 'melgan/layer16/stack/layer2/bias:0' shape=(48,) dtype=float32>, <tf.Variable 'melgan/layer16/stack/layer4/W:0' shape=(1, 48, 48) dtype=float32>, <tf.Variable 'melgan/layer16/stack/layer4/bias:0' shape=(48,) dtype=float32>, <tf.Variable 'melgan/layer17/skip_layer/W:0' shape=(1, 48, 48) dtype=float32>, <tf.Variable 'melgan/layer17/skip_layer/bias:0' shape=(48,) dtype=float32>, <tf.Variable 'melgan/layer17/stack/layer2/W:0' shape=(3, 48, 48) dtype=float32>, <tf.Variable 'melgan/layer17/stack/layer2/bias:0' shape=(48,) dtype=float32>, <tf.Variable 'melgan/layer17/stack/layer4/W:0' shape=(1, 48, 48) dtype=float32>, <tf.Variable 'melgan/layer17/stack/layer4/bias:0' shape=(48,) dtype=float32>, <tf.Variable 'melgan/layer18/skip_layer/W:0' shape=(1, 48, 48) dtype=float32>, <tf.Variable 'melgan/layer18/skip_layer/bias:0' shape=(48,) dtype=float32>, <tf.Variable 'melgan/layer18/stack/layer2/W:0' shape=(3, 48, 48) dtype=float32>, <tf.Variable 'melgan/layer18/stack/layer2/bias:0' shape=(48,) dtype=float32>, <tf.Variable 'melgan/layer18/stack/layer4/W:0' shape=(1, 48, 48) dtype=float32>, <tf.Variable 'melgan/layer18/stack/layer4/bias:0' shape=(48,) dtype=float32>, <tf.Variable 'melgan/layer19/skip_layer/W:0' shape=(1, 48, 48) dtype=float32>, <tf.Variable 'melgan/layer19/skip_layer/bias:0' shape=(48,) dtype=float32>, <tf.Variable 'melgan/layer19/stack/layer2/W:0' shape=(3, 48, 48) dtype=float32>, <tf.Variable 'melgan/layer19/stack/layer2/bias:0' shape=(48,) dtype=float32>, <tf.Variable 'melgan/layer19/stack/layer4/W:0' shape=(1, 48, 48) dtype=float32>, <tf.Variable 'melgan/layer19/stack/layer4/bias:0' shape=(48,) dtype=float32>, <tf.Variable 'melgan/layer22/W:0' shape=(7, 48, 4) dtype=float32>, <tf.Variable 'melgan/layer22/bias:0' shape=(4,) dtype=float32>, <tf.Variable 'melgan/layer3/W_native_transposed_conv:0' shape=(10, 1, 192, 384) dtype=float32>, <tf.Variable 'melgan/layer3/bias:0' shape=(192,) dtype=float32>, <tf.Variable 'melgan/layer4/skip_layer/W:0' shape=(1, 192, 192) dtype=float32>, <tf.Variable 'melgan/layer4/skip_layer/bias:0' shape=(192,) dtype=float32>, <tf.Variable 'melgan/layer4/stack/layer2/W:0' shape=(3, 192, 192) dtype=float32>, <tf.Variable 'melgan/layer4/stack/layer2/bias:0' shape=(192,) dtype=float32>, <tf.Variable 'melgan/layer4/stack/layer4/W:0' shape=(1, 192, 192) dtype=float32>, <tf.Variable 'melgan/layer4/stack/layer4/bias:0' shape=(192,) dtype=float32>, <tf.Variable 'melgan/layer5/skip_layer/W:0' shape=(1, 192, 192) dtype=float32>, <tf.Variable 'melgan/layer5/skip_layer/bias:0' shape=(192,) dtype=float32>, <tf.Variable 'melgan/layer5/stack/layer2/W:0' shape=(3, 192, 192) dtype=float32>, <tf.Variable 'melgan/layer5/stack/layer2/bias:0' shape=(192,) dtype=float32>, <tf.Variable 'melgan/layer5/stack/layer4/W:0' shape=(1, 192, 192) dtype=float32>, <tf.Variable 'melgan/layer5/stack/layer4/bias:0' shape=(192,) dtype=float32>, <tf.Variable 'melgan/layer6/skip_layer/W:0' shape=(1, 192, 192) dtype=float32>, <tf.Variable 'melgan/layer6/skip_layer/bias:0' shape=(192,) dtype=float32>, <tf.Variable 'melgan/layer6/stack/layer2/W:0' shape=(3, 192, 192) dtype=float32>, <tf.Variable 'melgan/layer6/stack/layer2/bias:0' shape=(192,) dtype=float32>, <tf.Variable 'melgan/layer6/stack/layer4/W:0' shape=(1, 192, 192) dtype=float32>, <tf.Variable 'melgan/layer6/stack/layer4/bias:0' shape=(192,) dtype=float32>, <tf.Variable 'melgan/layer7/skip_layer/W:0' shape=(1, 192, 192) dtype=float32>, <tf.Variable 'melgan/layer7/skip_layer/bias:0' shape=(192,) dtype=float32>, <tf.Variable 'melgan/layer7/stack/layer2/W:0' shape=(3, 192, 192) dtype=float32>, <tf.Variable 'melgan/layer7/stack/layer2/bias:0' shape=(192,) dtype=float32>, <tf.Variable 'melgan/layer7/stack/layer4/W:0' shape=(1, 192, 192) dtype=float32>, <tf.Variable 'melgan/layer7/stack/layer4/bias:0' shape=(192,) dtype=float32>, <tf.Variable 'melgan/layer9/W_native_transposed_conv:0' shape=(10, 1, 96, 192) dtype=float32>, <tf.Variable 'melgan/layer9/bias:0' shape=(96,) dtype=float32>]
Saving TF checkpoint to '/var/folders/fk/mt9zfm3n2853v1pcy0q3y25h0000gp/T/tmputue4givtmp-returnn-tf-checkpoint/model'...
>>> Constructing RETURNN model, load TF checkpoint, run...
layer <network via _run_returnn_standalone>/'data' output: Data(name='data', shape=(80, None), time_dim_axis=2, feature_dim_axis=1, batch_shape_meta=[B,F|80,T|'time:var:extern_data:data'])
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5903935888),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5903936144),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5903933648),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5903936272)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5903936400)]
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5904008016),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904008272),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5904007696),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5904008400)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904008528)]
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5904051344),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904051600),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5904048464),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5904051728)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904051856)]
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5904160208),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904160528),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5904159376),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5904161232)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904161104)]
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5904250064),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904250320),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5904249296),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5904250448)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904250576)]
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5904301584),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904301840),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5904299344),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5904301968)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904302032)]
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5904406416),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904406672),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5904405072),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5904406800)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904406928)]
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5904498832),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904499984),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5904499280),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5904500240)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904500368)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5904589264),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904589520),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5904588944),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5904589648)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904589776)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5904657168),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904657424),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5904654928),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5904657552)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904657680)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5904741520),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904741776),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5904740432),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5904741904)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904742032)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5904743248),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904831056),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5904831248),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5904831184)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904831376)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5904883280),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904883536),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5904881040),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5904883664)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904941200)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5904251600),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904251792),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5904251984),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5904250064)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5904250512)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5905060624),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905060880),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5905059984),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5905061008)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905061136)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5905150416),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905150672),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5905150224),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5905150800)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905150928)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5905230672),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905230928),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5905228432),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5905231056)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905231184)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5905323216),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905323472),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5905322448),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5905323600)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905323728)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5905274960),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905275088),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5905274576),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5905275216)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905274768)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5905464208),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905464784),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5905462288),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5905464912)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905465040)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5905564816),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905565072),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5905564048),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5905565200)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905565328)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5905674320),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905674448),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5905674384),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5905674576)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905674704)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5905712272),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905712656),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5905712016),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5905713552)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905712976)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5905811472),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905811728),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5905809552),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5905811856)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905811984)]
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5906002064),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5906002192),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5906002128),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5906002320)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5906002448)]
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5906032720),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5906034000),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5906032016),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5906034256)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5906034384)]
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5906127312),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5906127568),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5906125072),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5906127696)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5906127824)]
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5906228048),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5906228304),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5906227728),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5906228432)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5906228560)]
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5905937872),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905937936),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5905937552),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5905938064)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905892496)]
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5750462672),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5750461200),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5750463952),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5717892880)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5717893072)]
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5869215632),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5869215696),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5869168592),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5869167184)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5869167376)]
get_common_data(
[Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]),
Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5878956624),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5878956112),
DimensionTag(kind='feature', description='feature:output_output', dimension=192, id=5878956816),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5878959184)],
largest source
(Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5878956368)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5874784272),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5874695184),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5874783632),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5874649488)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5874650960)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5878663248),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5878708176),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5878664528),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5866757520)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5866754832)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5868640208),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5868641680),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5868641424),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5868640976)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5868551504)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5866603920),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5855352720),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5855351824),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5855352272)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5855353680)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5854631120),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5854630480),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5854630672),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5854631888)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5854565584)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5866127888),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5854297680),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5854295184),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5854295632)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5854297808)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5851690320),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5851663440),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5851661840),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5851663248)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5851662352)]
get_common_data(
[Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]),
Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5828863056),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5828863568),
DimensionTag(kind='feature', description='feature:output_output', dimension=96, id=5828864976),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5828863824)],
largest source
(Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5828864080)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5926735376),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5926733392),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5926735440),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5926732816)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5926732304)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5834235024),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5834234128),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5834235536),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5851924560)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5805873232)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5867604624),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5717876880),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5717877648),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5717878416)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5717879696)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5844809872),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5844754128),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5844753808),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5844753232)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5844754000)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5844435728),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5844437520),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5844438800),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5844435024)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5844438160)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5845135760),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5845136144),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5845137936),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5844703376)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5844703888)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5905863120),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905864144),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5905862992),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5905863952)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5905864656)]
get_common_data(
[Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]),
Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])]),
dim tags
[DimensionTag(kind='batch', description='batch:output_output', id=5750452048),
DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5750449744),
DimensionTag(kind='feature', description='feature:output_output', dimension=48, id=5750450000),
DimensionTag(kind='spatial', description='time:var-unk:skip_layer_output', id=5750450960)],
largest source
(Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48]))
has incomplete dim tag info:
[DimensionTag(kind='spatial', description='time:var-unk:output_output', id=5750449616)]
layer <network via _run_returnn_standalone>/'melgan' output: Data(name='output_output', shape=(None, 4), batch_shape_meta=[B,T|?,F|4])
layer <network via _run_returnn_standalone>/melgan:subnet/'data' output: Data(name='data', shape=(80, None), time_dim_axis=2, feature_dim_axis=1, batch_shape_meta=[B,F|80,T|'time:var:extern_data:data'])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer0' output: Data(name='layer0_output', shape=(80, None), time_dim_axis=2, feature_dim_axis=1, batch_shape_meta=[B,F|80,T|'time:var:extern_data:data'])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer1' output: Data(name='layer1_output', shape=(None, 384), batch_shape_meta=[B,T|?,F|384])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer2' output: Data(name='layer2_output', shape=(None, 384), batch_shape_meta=[B,T|'time:var:extern_data:data',F|384])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer3' output: Data(name='layer3_output', shape=(None, 192), batch_shape_meta=[B,T|?,F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer4' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer4:subnet/'data' output: Data(name='layer3_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer4:subnet/'stack' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer4:subnet/stack:subnet/'data' output: Data(name='layer3_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer4:subnet/stack:subnet/'layer0' output: Data(name='layer0_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer4:subnet/stack:subnet/'layer1' output: Data(name='layer1_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer4:subnet/stack:subnet/'layer2' output: Data(name='layer2_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer4/stack/layer1',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer4:subnet/stack:subnet/'layer3' output: Data(name='layer3_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer4:subnet/stack:subnet/'layer4' output: Data(name='layer4_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer4:subnet/stack:subnet/'output' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer4:subnet/'skip_layer' output: Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer4:subnet/'add' output: Data(name='add_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer4:subnet/'output' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer5' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer5:subnet/'data' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer5:subnet/'stack' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer5:subnet/stack:subnet/'data' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer5:subnet/stack:subnet/'layer0' output: Data(name='layer0_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer5:subnet/stack:subnet/'layer1' output: Data(name='layer1_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer5:subnet/stack:subnet/'layer2' output: Data(name='layer2_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer5/stack/layer1',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer5:subnet/stack:subnet/'layer3' output: Data(name='layer3_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer5:subnet/stack:subnet/'layer4' output: Data(name='layer4_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer5:subnet/stack:subnet/'output' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer5:subnet/'skip_layer' output: Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer5:subnet/'add' output: Data(name='add_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer5:subnet/'output' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer6' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer6:subnet/'data' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer6:subnet/'stack' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer6:subnet/stack:subnet/'data' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer6:subnet/stack:subnet/'layer0' output: Data(name='layer0_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer6:subnet/stack:subnet/'layer1' output: Data(name='layer1_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer6:subnet/stack:subnet/'layer2' output: Data(name='layer2_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer6/stack/layer1',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer6:subnet/stack:subnet/'layer3' output: Data(name='layer3_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer6:subnet/stack:subnet/'layer4' output: Data(name='layer4_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer6:subnet/stack:subnet/'output' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer6:subnet/'skip_layer' output: Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer6:subnet/'add' output: Data(name='add_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer6:subnet/'output' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer7' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer7:subnet/'data' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer7:subnet/'stack' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer7:subnet/stack:subnet/'data' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer7:subnet/stack:subnet/'layer0' output: Data(name='layer0_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer7:subnet/stack:subnet/'layer1' output: Data(name='layer1_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer7:subnet/stack:subnet/'layer2' output: Data(name='layer2_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer7/stack/layer1',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer7:subnet/stack:subnet/'layer3' output: Data(name='layer3_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer7:subnet/stack:subnet/'layer4' output: Data(name='layer4_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer7:subnet/stack:subnet/'output' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer7:subnet/'skip_layer' output: Data(name='skip_layer_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer7:subnet/'add' output: Data(name='add_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/layer7:subnet/'output' output: Data(name='output_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer8' output: Data(name='layer8_output', shape=(None, 192), batch_shape_meta=[B,T|'spatial:0:melgan/layer3',F|192])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer9' output: Data(name='layer9_output', shape=(None, 96), batch_shape_meta=[B,T|?,F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer10' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer10:subnet/'data' output: Data(name='layer9_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer10:subnet/'stack' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer10:subnet/stack:subnet/'data' output: Data(name='layer9_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer10:subnet/stack:subnet/'layer0' output: Data(name='layer0_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer10:subnet/stack:subnet/'layer1' output: Data(name='layer1_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer10:subnet/stack:subnet/'layer2' output: Data(name='layer2_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer10/stack/layer1',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer10:subnet/stack:subnet/'layer3' output: Data(name='layer3_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer10:subnet/stack:subnet/'layer4' output: Data(name='layer4_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer10:subnet/stack:subnet/'output' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer10:subnet/'skip_layer' output: Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer10:subnet/'add' output: Data(name='add_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer10:subnet/'output' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer11' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer11:subnet/'data' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer11:subnet/'stack' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer11:subnet/stack:subnet/'data' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer11:subnet/stack:subnet/'layer0' output: Data(name='layer0_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer11:subnet/stack:subnet/'layer1' output: Data(name='layer1_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer11:subnet/stack:subnet/'layer2' output: Data(name='layer2_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer11/stack/layer1',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer11:subnet/stack:subnet/'layer3' output: Data(name='layer3_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer11:subnet/stack:subnet/'layer4' output: Data(name='layer4_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer11:subnet/stack:subnet/'output' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer11:subnet/'skip_layer' output: Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer11:subnet/'add' output: Data(name='add_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer11:subnet/'output' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer12' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer12:subnet/'data' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer12:subnet/'stack' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer12:subnet/stack:subnet/'data' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer12:subnet/stack:subnet/'layer0' output: Data(name='layer0_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer12:subnet/stack:subnet/'layer1' output: Data(name='layer1_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer12:subnet/stack:subnet/'layer2' output: Data(name='layer2_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer12/stack/layer1',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer12:subnet/stack:subnet/'layer3' output: Data(name='layer3_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer12:subnet/stack:subnet/'layer4' output: Data(name='layer4_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer12:subnet/stack:subnet/'output' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer12:subnet/'skip_layer' output: Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer12:subnet/'add' output: Data(name='add_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer12:subnet/'output' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer13' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer13:subnet/'data' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer13:subnet/'stack' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer13:subnet/stack:subnet/'data' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer13:subnet/stack:subnet/'layer0' output: Data(name='layer0_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer13:subnet/stack:subnet/'layer1' output: Data(name='layer1_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer13:subnet/stack:subnet/'layer2' output: Data(name='layer2_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer13/stack/layer1',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer13:subnet/stack:subnet/'layer3' output: Data(name='layer3_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer13:subnet/stack:subnet/'layer4' output: Data(name='layer4_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer13:subnet/stack:subnet/'output' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer13:subnet/'skip_layer' output: Data(name='skip_layer_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer13:subnet/'add' output: Data(name='add_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/layer13:subnet/'output' output: Data(name='output_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer14' output: Data(name='layer14_output', shape=(None, 96), batch_shape_meta=[B,T|'spatial:0:melgan/layer9',F|96])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer15' output: Data(name='layer15_output', shape=(None, 48), batch_shape_meta=[B,T|?,F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer16' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer16:subnet/'data' output: Data(name='layer15_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer16:subnet/'stack' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer16:subnet/stack:subnet/'data' output: Data(name='layer15_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer16:subnet/stack:subnet/'layer0' output: Data(name='layer0_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer16:subnet/stack:subnet/'layer1' output: Data(name='layer1_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer16:subnet/stack:subnet/'layer2' output: Data(name='layer2_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer16/stack/layer1',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer16:subnet/stack:subnet/'layer3' output: Data(name='layer3_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer16:subnet/stack:subnet/'layer4' output: Data(name='layer4_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer16:subnet/stack:subnet/'output' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer16:subnet/'skip_layer' output: Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer16:subnet/'add' output: Data(name='add_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer16:subnet/'output' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer17' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer17:subnet/'data' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer17:subnet/'stack' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer17:subnet/stack:subnet/'data' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer17:subnet/stack:subnet/'layer0' output: Data(name='layer0_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer17:subnet/stack:subnet/'layer1' output: Data(name='layer1_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer17:subnet/stack:subnet/'layer2' output: Data(name='layer2_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer17/stack/layer1',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer17:subnet/stack:subnet/'layer3' output: Data(name='layer3_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer17:subnet/stack:subnet/'layer4' output: Data(name='layer4_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer17:subnet/stack:subnet/'output' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer17:subnet/'skip_layer' output: Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer17:subnet/'add' output: Data(name='add_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer17:subnet/'output' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer18' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer18:subnet/'data' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer18:subnet/'stack' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer18:subnet/stack:subnet/'data' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer18:subnet/stack:subnet/'layer0' output: Data(name='layer0_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer18:subnet/stack:subnet/'layer1' output: Data(name='layer1_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer18:subnet/stack:subnet/'layer2' output: Data(name='layer2_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer18/stack/layer1',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer18:subnet/stack:subnet/'layer3' output: Data(name='layer3_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer18:subnet/stack:subnet/'layer4' output: Data(name='layer4_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer18:subnet/stack:subnet/'output' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer18:subnet/'skip_layer' output: Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer18:subnet/'add' output: Data(name='add_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer18:subnet/'output' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer19' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer19:subnet/'data' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer19:subnet/'stack' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer19:subnet/stack:subnet/'data' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer19:subnet/stack:subnet/'layer0' output: Data(name='layer0_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer19:subnet/stack:subnet/'layer1' output: Data(name='layer1_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer19:subnet/stack:subnet/'layer2' output: Data(name='layer2_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer19/stack/layer1',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer19:subnet/stack:subnet/'layer3' output: Data(name='layer3_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer19:subnet/stack:subnet/'layer4' output: Data(name='layer4_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer19:subnet/stack:subnet/'output' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer19:subnet/'skip_layer' output: Data(name='skip_layer_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer19:subnet/'add' output: Data(name='add_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/layer19:subnet/'output' output: Data(name='output_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer20' output: Data(name='layer20_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer21' output: Data(name='layer21_output', shape=(None, 48), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|48])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer22' output: Data(name='layer22_output', shape=(None, 4), batch_shape_meta=[B,T|'spatial:0:melgan/layer21',F|4])
layer <network via _run_returnn_standalone>/melgan:subnet/'layer23' output: Data(name='layer23_output', shape=(None, 4), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|4])
layer <network via _run_returnn_standalone>/melgan:subnet/'output' output: Data(name='output_output', shape=(None, 4), batch_shape_meta=[B,T|'spatial:0:melgan/layer15',F|4])
layer <network via _run_returnn_standalone>/'PQMF_updown_filter' output: Data(name='PQMF_updown_filter_const', shape=(4, 4, 4), batch_dim_axis=None, time_dim_axis=None, batch_shape_meta=[4,4,F|4])
layer <network via _run_returnn_standalone>/'PQMF_Cast_unnamed_const' output: Data(name='PQMF_Cast_unnamed_const_const', shape=(), dtype='int32', batch_dim_axis=None, time_dim_axis=None, batch_shape_meta=[])
layer <network via _run_returnn_standalone>/'PQMF_Cast' output: Data(name='PQMF_Cast_output', shape=(), batch_dim_axis=None, time_dim_axis=None, batch_shape_meta=[])
layer <network via _run_returnn_standalone>/'PQMF_mul' output: Data(name='PQMF_mul_output', shape=(4, 4, 4), batch_dim_axis=None, time_dim_axis=None, batch_shape_meta=[4,4,F|4])
layer <network via _run_returnn_standalone>/'PQMF_FunctionalConvTransposed1d' output: Data(name='PQMF_FunctionalConvTransposed1d_output', shape=(None, 4), batch_shape_meta=[B,T|?,F|4])
layer <network via _run_returnn_standalone>/'pad_fn' output: Data(name='pad_fn_output', shape=(None, 4), batch_shape_meta=[B,T|'spatial:0:PQMF_FunctionalConvTransposed1d',F|4])
layer <network via _run_returnn_standalone>/'PQMF_synthesis_filter' output: Data(name='PQMF_synthesis_filter_const', shape=(1, 4, 63), batch_dim_axis=None, time_dim_axis=None, batch_shape_meta=[1,4,F|63])
layer <network via _run_returnn_standalone>/'PQMF_FunctionalConv1d' output: Data(name='PQMF_FunctionalConv1d_output', shape=(None, 1), batch_shape_meta=[B,T|'spatial:0:pad_fn',F|1])
layer <network via _run_returnn_standalone>/'output' output: Data(name='output_output', shape=(None, 1), batch_shape_meta=[B,T|'spatial:0:PQMF_FunctionalConvTransposed1d',F|1])
Output shape: (1, 16000, 1)
>>>> Looks good!
Wrote out.wav.
Process finished with exit code 0
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment