yolov8模型转换：

if [ ! -f /app/project/best.onnx ];then

    echo "export model begin..."

    cd $PROJ_ROOT_PATH/export_model/ultralytics/

    python3.8 setup.py install

    yolo task=detect mode=export model=/app/project/best.pt format=onnx source=${PROJ_ROOT_PATH}/export_model/ultralytics/assets/bus.jpg opset=11 dynamic=true simplify=true

    echo "export model end..."

fi

报错：

ONNX: starting export with onnx 1.12.0 opset 11...

/usr/local/lib/python3.8/dist-packages/torch/tensor.py:590: RuntimeWarning: Iterating over a tensor might cause the trace to be incorrect. Passing a tensor of different shape won't change the number of iterations executed (and might lead to errors or silently give incorrect results).

  warnings.warn('Iterating over a tensor might cause the trace to be incorrect. '

ERROR ❌ ONNX: export failure 0.5s: Exporting the operator silu to ONNX opset version 11 is not supported. Please open a bug to request ONNX export support for the missing operator.

Traceback (most recent call last):

  File "/usr/local/bin/yolo", line 11, in <module>

    load_entry_point('ultralytics==8.3.133', 'console_ s', 'yolo')()

  File "/app/magicmind_cloud/buildin/cv/detection/yolov8_pytorch/export_model/ultralytics/ultralytics/cfg/__init__.py", line 981, in entrypoint

    getattr(model, mode)(**overrides)  # default args from model

  File "/app/magicmind_cloud/buildin/cv/detection/yolov8_pytorch/export_model/ultralytics/ultralytics/engine/model.py", line 730, in export

    return Exporter(overrides=args, _callbacks=self.callbacks)(model=self.model)

  File "/app/magicmind_cloud/buildin/cv/detection/yolov8_pytorch/export_model/ultralytics/ultralytics/engine/exporter.py", line 461, in __call__

    f[2], _ = self.export_onnx()

  File "/app/magicmind_cloud/buildin/cv/detection/yolov8_pytorch/export_model/ultralytics/ultralytics/engine/exporter.py", line 197, in outer_func

    raise e

  File "/app/magicmind_cloud/buildin/cv/detection/yolov8_pytorch/export_model/ultralytics/ultralytics/engine/exporter.py", line 192, in outer_func

    f, model = inner_func(*args, **kwargs)

  File "/app/magicmind_cloud/buildin/cv/detection/yolov8_pytorch/export_model/ultralytics/ultralytics/engine/exporter.py", line 582, in export_onnx

    export_onnx(

  File "/app/magicmind_cloud/buildin/cv/detection/yolov8_pytorch/export_model/ultralytics/ultralytics/utils/export.py", line 35, in export_onnx

    torch.onnx.export(

  File "/usr/local/lib/python3.8/dist-packages/torch/onnx/__init__.py", line 225, in export

    return utils.export(model, args, f, export_params, verbose, training,

  File "/usr/local/lib/python3.8/dist-packages/torch/onnx/utils.py", line 85, in export

    _export(model, args, f, export_params, verbose, training, input_names, output_names,

  File "/usr/local/lib/python3.8/dist-packages/torch/onnx/utils.py", line 632, in _export

    _model_to_graph(model, args, verbose, input_names,

  File "/usr/local/lib/python3.8/dist-packages/torch/onnx/utils.py", line 417, in _model_to_graph

    graph = _optimize_graph(graph, operator_export_type,

  File "/usr/local/lib/python3.8/dist-packages/torch/onnx/utils.py", line 203, in _optimize_graph

    graph = torch._C._jit_pass_onnx(graph, operator_export_type)

  File "/usr/local/lib/python3.8/dist-packages/torch/onnx/__init__.py", line 263, in _run_symbolic_function

    return utils._run_symbolic_function(*args, **kwargs)

  File "/usr/local/lib/python3.8/dist-packages/torch/onnx/utils.py", line 930, in _run_symbolic_function

    symbolic_fn = _find_symbolic_in_registry(domain, op_name, opset_version, operator_export_type)

  File "/usr/local/lib/python3.8/dist-packages/torch/onnx/utils.py", line 888, in _find_symbolic_in_registry

    return sym_registry.get_registered_op(op_name, domain, opset_version)

  File "/usr/local/lib/python3.8/dist-packages/torch/onnx/symbolic_registry.py", line 111, in get_registered_op

    raise RuntimeError(msg)

RuntimeError: Exporting the operator silu to ONNX opset version 11 is not supported. Please open a bug to request ONNX export support for the missing operator.

官方gitee文档里解释如下：

由于 MagicMind 最高支持 PyTorch 1.6.0 版本，此版本没有 SiLU 函数，所以要在环境中修改代码如下：

vim /usr/lib/python3.7/site-packages/torch/nn/modules/activation.py#
进行如下修改return F.silu(input, inplace=self.inplace) 替换为 return input * torch.sigmoid(input)

但是我的activation.py并没有

return F.silu(input, inplace=self.inplace)