stacktrace.txt

 Unhandled exception <class 'RuntimeError'> in thread <_MainThread(MainThread, started 140574602995520)>, proc 71253.

Thread current, main, <_MainThread(MainThread, started 140574602995520)>:
(Excluded thread.)

That were all threads.
EXCEPTION
Traceback (most recent call last):
  File "/home/nbeneitez/Documentos/work/repos/returnn_pytorch/tools/torch_export_to_onnx.py", line 209, in <module>
    line: main()
    locals:
      main = <local> <function main at 0x7fd950a55a20>
  File "/home/nbeneitez/Documentos/work/repos/returnn_pytorch/tools/torch_export_to_onnx.py", line 197, in main
    line: export_func(
              pt_model_fwd,
              (extern_data_raw, {}),
              f=args.out_onnx_filename,
              verbose=True,
              input_names=list(extern_data_raw.keys()),
              output_names=model_outputs_raw_keys,
              dynamic_axes=dynamic_axes,
          )
    locals:
      export_func = <local> <function export at 0x7fd95339fb50>
      pt_model_fwd = <local> ForwardModulePT(
                               (model): Model(
                                 (layers): Sequential(
                                   (0): Conv1d(9, 50, kernel_size=(5,), stride=(1,), padding=(2,))
                                   (1): ReLU()
                                   (2): Conv1d(50, 100, kernel_size=(5,), stride=(1,), padding=(2,))
                                   (3): ReLU()
                                   (4): Conv1d(100, 2, kernel_size=(5,), stride=(1,), p...
      extern_data_raw = <local> {'data': tensor([[[-0.2342, -0.2341,  1.5792,  0.7674, -0.4695,  0.5426, -0.4634,
                                          -0.4657,  0.2420],
                                         [-1.9133, -1.7249, -0.5623, -1.0128,  0.3142, -0.9080, -1.4123,
                                           1.4656, -0.2258],
                                         [ 0.0675, -1.4247, -0.5444,  0.1109, -1.1510,  0.3757, -0.6006,
                                          -..., len = 6
      f = <not found>
      args = <local> Namespace(config='demos/demo-torch.config', checkpoint='/tmp/nbeneitez/returnn/demos/demo-torch/model.005.pt', out_onnx_filename='/tmp/nbeneitez/returnn/demos/demo-torch/model.005.onnx', verbosity=4, device='cpu')
      args.out_onnx_filename = <local> '/tmp/nbeneitez/returnn/demos/demo-torch/model.005.onnx', len = 54
      verbose = <not found>
      input_names = <not found>
      list = <builtin> <class 'list'>
      extern_data_raw.keys = <local> <built-in method keys of dict object at 0x7fd9522d12c0>
      output_names = <not found>
      model_outputs_raw_keys = <local> ['output', 'output:size0', 'output:size1'], _[0]: {len = 6}
      dynamic_axes = <local> {'data': {0: 'global batch', 1: 'time:var-unk:data'}, 'data:size0': {}, 'data:size1': {0: 'global batch'}, 'classes': {0: 'global batch', 1: 'time:var-unk:classes'}, 'classes:size0': {}, 'classes:size1': {0: 'global batch'}, 'output': {0: 'global batch', 1: 'time:var-unk:output'}, 'output:size0':..., len = 9
  File "/home/nbeneitez/.venv/returnn_pytorch/lib/python3.10/site-packages/torch/onnx/utils.py", line 506, in export
    line: _export(
              model,
              args,
              f,
              export_params,
              verbose,
              training,
              input_names,
              output_names,
              operator_export_type=operator_export_type,
              opset_version=opset_version,
              do_constant_folding=do_constant_folding,
              dynamic_axes=dynamic_axes,
              keep_initializers_as_inputs=keep_initializers_as_inputs,
              custom_opsets=custom_opsets,
              export_modules_as_functions=export_modules_as_functions,
          )
    locals:
      _export = <global> <function _export at 0x7fd9533a0a60>
      model = <local> ForwardModulePT(
                        (model): Model(
                          (layers): Sequential(
                            (0): Conv1d(9, 50, kernel_size=(5,), stride=(1,), padding=(2,))
                            (1): ReLU()
                            (2): Conv1d(50, 100, kernel_size=(5,), stride=(1,), padding=(2,))
                            (3): ReLU()
                            (4): Conv1d(100, 2, kernel_size=(5,), stride=(1,), p...
      args = <local> ({'data': tensor([[[-0.2342, -0.2341,  1.5792,  0.7674, -0.4695,  0.5426, -0.4634,
                               -0.4657,  0.2420],
                              [-1.9133, -1.7249, -0.5623, -1.0128,  0.3142, -0.9080, -1.4123,
                                1.4656, -0.2258],
                              [ 0.0675, -1.4247, -0.5444,  0.1109, -1.1510,  0.3757, -0.6006,
                               ..., _[0]: {len = 6}
      f = <local> '/tmp/nbeneitez/returnn/demos/demo-torch/model.005.onnx', len = 54
      export_params = <local> True
      verbose = <local> True
      training = <local> <TrainingMode.EVAL: 0>
      input_names = <local> ['data', 'data:size0', 'data:size1', 'classes', 'classes:size0', 'classes:size1'], len = 6
      output_names = <local> ['output', 'output:size0', 'output:size1'], _[0]: {len = 6}
      operator_export_type = <local> <OperatorExportTypes.ONNX: 0>
      opset_version = <local> None
      do_constant_folding = <local> True
      dynamic_axes = <local> {'data': {0: 'global batch', 1: 'time:var-unk:data'}, 'data:size0': {}, 'data:size1': {0: 'global batch'}, 'classes': {0: 'global batch', 1: 'time:var-unk:classes'}, 'classes:size0': {}, 'classes:size1': {0: 'global batch'}, 'output': {0: 'global batch', 1: 'time:var-unk:output'}, 'output:size0':..., len = 9
      keep_initializers_as_inputs = <local> None
      custom_opsets = <local> None
      export_modules_as_functions = <local> False
  File "/home/nbeneitez/.venv/returnn_pytorch/lib/python3.10/site-packages/torch/onnx/utils.py", line 1548, in _export
    line: graph, params_dict, torch_out = _model_to_graph(
              model,
              args,
              verbose,
              input_names,
              output_names,
              operator_export_type,
              val_do_constant_folding,
              fixed_batch_size=fixed_batch_size,
              training=training,
              dynamic_axes=dynamic_axes,
          )
    locals:
      graph = <not found>
      params_dict = <not found>
      torch_out = <not found>
      _model_to_graph = <global> <function _model_to_graph at 0x7fd9533a0700>
      model = <local> ForwardModulePT(
                        (model): Model(
                          (layers): Sequential(
                            (0): Conv1d(9, 50, kernel_size=(5,), stride=(1,), padding=(2,))
                            (1): ReLU()
                            (2): Conv1d(50, 100, kernel_size=(5,), stride=(1,), padding=(2,))
                            (3): ReLU()
                            (4): Conv1d(100, 2, kernel_size=(5,), stride=(1,), p...
      args = <local> ({'data': tensor([[[-0.2342, -0.2341,  1.5792,  0.7674, -0.4695,  0.5426, -0.4634,
                               -0.4657,  0.2420],
                              [-1.9133, -1.7249, -0.5623, -1.0128,  0.3142, -0.9080, -1.4123,
                                1.4656, -0.2258],
                              [ 0.0675, -1.4247, -0.5444,  0.1109, -1.1510,  0.3757, -0.6006,
                               ..., _[0]: {len = 6}
      verbose = <local> True
      input_names = <local> ['data', 'data:size0', 'data:size1', 'classes', 'classes:size0', 'classes:size1'], len = 6
      output_names = <local> ['output', 'output:size0', 'output:size1'], _[0]: {len = 6}
      operator_export_type = <local> <OperatorExportTypes.ONNX: 0>
      val_do_constant_folding = <local> True
      fixed_batch_size = <local> False
      training = <local> <TrainingMode.EVAL: 0>
      dynamic_axes = <local> {'data': {0: 'global batch', 1: 'time:var-unk:data'}, 'data:size0': {}, 'data:size1': {0: 'global batch'}, 'classes': {0: 'global batch', 1: 'time:var-unk:classes'}, 'classes:size0': {}, 'classes:size1': {0: 'global batch'}, 'output': {0: 'global batch', 1: 'time:var-unk:output'}, 'output:size0':..., len = 9
  File "/home/nbeneitez/.venv/returnn_pytorch/lib/python3.10/site-packages/torch/onnx/utils.py", line 1169, in _model_to_graph
    line: _set_input_and_output_names(graph, input_names, output_names)
    locals:
      _set_input_and_output_names = <global> <function _set_input_and_output_names at 0x7fd9533a0b80>
      graph = <local> graph(%data : Float(*, *, 9, strides=[144, 9, 1], requires_grad=0, device=cpu),
                            %data:size0 : Int(requires_grad=0, device=cpu),
                            %data:size1 : Int(*, strides=[1], requires_grad=0, device=cpu),
                            %classes : Int(*, *, strides=[7, 1], requires_grad=0, device=cpu),
                            %classes:size0...
      input_names = <local> ['data', 'data:size0', 'data:size1', 'classes', 'classes:size0', 'classes:size1'], len = 6
      output_names = <local> ['output', 'output:size0', 'output:size1'], _[0]: {len = 6}
  File "/home/nbeneitez/.venv/returnn_pytorch/lib/python3.10/site-packages/torch/onnx/utils.py", line 1690, in _set_input_and_output_names
    line: set_names(list(graph.outputs()), output_names, "output")
    locals:
      set_names = <local> <function _set_input_and_output_names.<locals>.set_names at 0x7fd9508779a0>
      list = <builtin> <class 'list'>
      graph = <local> graph(%data : Float(*, *, 9, strides=[144, 9, 1], requires_grad=0, device=cpu),
                            %data:size0 : Int(requires_grad=0, device=cpu),
                            %data:size1 : Int(*, strides=[1], requires_grad=0, device=cpu),
                            %classes : Int(*, *, strides=[7, 1], requires_grad=0, device=cpu),
                            %classes:size0...
      graph.outputs = <local> <bound method PyCapsule.outputs of graph(%data : Float(*, *, 9, strides=[144, 9, 1], requires_grad=0, device=cpu),
                                    %data:size0 : Int(requires_grad=0, device=cpu),
                                    %data:size1 : Int(*, strides=[1], requires_grad=0, device=cpu),
                                    %classes : Int(*, *, strides=[7, 1], requires_grad=0...
      output_names = <local> ['output', 'output:size0', 'output:size1'], _[0]: {len = 6}
  File "/home/nbeneitez/.venv/returnn_pytorch/lib/python3.10/site-packages/torch/onnx/utils.py", line 1667, in _set_input_and_output_names.<locals>.set_names
    line: raise RuntimeError(
              "number of %s names provided (%d) exceeded number of %ss (%d)"
              % (descriptor, len(name_list), descriptor, len(node_list))
          )
    locals:
      RuntimeError = <builtin> <class 'RuntimeError'>
      descriptor = <local> 'output', len = 6
      len = <builtin> <built-in function len>
      name_list = <local> ['output', 'output:size0', 'output:size1'], _[0]: {len = 6}
      node_list = <local> [raw_tensor defined in (%raw_tensor : Float(*, *, 2, strides=[32, 1, 16], requires_grad=1, device=cpu) = onnx::Transpose[perm=[0, 2, 1]](%17), scope: __returnn_config__.Model::model # demos/demo-torch.config:53:0
                          ), 19 defined in (%19 : Int(requires_grad=0, device=cpu) = onnx::Identity(%classes:s...
RuntimeError: number of output names provided (3) exceeded number of outputs (2)

Process finished with exit code 1

rwth-i6/returnn#1333 (comment)