无法将 Pytorch 转换为 ONNX

Question

尝试使用 ONNX 转换 this pytorch 模型时出现此错误。我在 github 上搜索过，这个错误之前在 1.1.0 版本中出现过，但显然已得到纠正。现在我在火炬 1.4.0 上。 (python 3.6.9)，我看到了这个错误。

File "/usr/local/lib/python3.6/dist-packages/torch/onnx/init.py", line 148, in export
strip_doc_string, dynamic_axes, keep_initializers_as_inputs)
File "/usr/local/lib/python3.6/dist-packages/torch/onnx/utils.py", line 66, in export
dynamic_axes=dynamic_axes, keep_initializers_as_inputs=keep_initializers_as_inputs)
File "/usr/local/lib/python3.6/dist-packages/torch/onnx/utils.py", line 416, in _export
fixed_batch_size=fixed_batch_size)
File "/usr/local/lib/python3.6/dist-packages/torch/onnx/utils.py", line 296, in _model_to_graph
fixed_batch_size=fixed_batch_size, params_dict=params_dict)
File "/usr/local/lib/python3.6/dist-packages/torch/onnx/utils.py", line 135, in _optimize_graph
graph = torch._C._jit_pass_onnx(graph, operator_export_type)
File "/usr/local/lib/python3.6/dist-packages/torch/onnx/init.py", line 179, in _run_symbolic_function
return utils._run_symbolic_function(*args, **kwargs)
File "/usr/local/lib/python3.6/dist-packages/torch/onnx/utils.py", line 657, in _run_symbolic_function
return op_fn(g, *inputs, **attrs)
File "/usr/local/lib/python3.6/dist-packages/torch/onnx/symbolic_helper.py", line 128, in wrapper
args = [_parse_arg(arg, arg_desc) for arg, arg_desc in zip(args, arg_descriptors)]
File "/usr/local/lib/python3.6/dist-packages/torch/onnx/symbolic_helper.py", line 128, in
args = [_parse_arg(arg, arg_desc) for arg, arg_desc in zip(args, arg_descriptors)]
File "/usr/local/lib/python3.6/dist-packages/torch/onnx/symbolic_helper.py", line 81, in _parse_arg
"', since it's not constant, please try to make "
RuntimeError: Failed to export an ONNX attribute 'onnx::Gather', since it's not constant, please try to make things (e.g., kernel size) static if possible

如何解决？我也尝试了最新的夜间构建，同样的错误出现了。

我的代码：

from model import BiSeNet
import torch.onnx
import torch

net = BiSeNet(19)
net.cuda()
net.load_state_dict(torch.load('/content/drive/My Drive/Collab/fp/res/cp/79999_iter.pth'))
net.eval()

dummy = torch.rand(1,3,512,512).cuda()
torch.onnx.export(net, dummy, "Model.onnx", input_names=["image"], output_names=["output"])

我在引发运行时错误之前将 print (v.node ()) 添加到 symbolic_helper.py 以查看导致错误的原因。

这是输出：%595 : Long() = onnx::Gather[axis=0](%592, %594) # /content/drive/My Drive/Collab/fp/model.py:111:0

model.py 中 111 中的那一行是：avg = F.avg_pool2d(feat32, feat32.size()[2:])

这个source表示pytorch中的tensor.size方法无法被onnx识别，需要修改成常量。

Answer 1

以前使用导出时也遇到过类似的错误

torch.onnx.export(model, x, ONNX_FILE_PATH)

我通过像这样指定opset_version 来修复它：

torch.onnx.export(model, x, ONNX_FILE_PATH, opset_version = 11)

Answer 2

你可以：

print(feat32.size()[2:])

并替换：

F.avg_pool2d(feat32, feat32.size()[2:]) 

与：

F.avg_pool2d(feat32, your_print_contant_result)

Answer 3

我也遇到过同样的问题。

我的情况不支持 F.adaptive_avg_pool2d。您必须尝试其他操作。

我希望这会有所帮助。

谢谢

Answer 4

更改实例

x = F.avg_pool2d(x, x.shape[2:])

到

x_shape = [int(s) for s in x.shape[2:]]
x = F.avg_pool2d(x, x_shape)

这样avg_pool2d 的输入形状是恒定的[k,k]，而不是torch.Size([k, k]) 提到的here。

Answer 5

对于那些在 Google 搜索中遇到此问题并获得 Unable to cast from non-held to held instance (T& to Holder) (compile in debug mode for type information) 的用户，请尝试像这样添加 operator_export_type=torch.onnx.OperatorExportTypes.ONNX_ATEN_FALLBACK (as mentioned here)：

torch.onnx.export(model, input, "output-name.onnx", export_params=True, opset_version=12, operator_export_type=torch.onnx.OperatorExportTypes.ONNX_ATEN_FALLBACK)

这解决了我的情况下的“持有实例”问题。