How to edit onnx(converted from tf with tf2onnx.convert --large_model) larger than 2GB with onnx_graphsurgeon?

[MsDidin]

Description

I have a onnx model converted form tf with tf2onnx tool. when converted from tf, the model is large than 2GB so I have used the --large_model option

when I use onnx_graphsurgeon to get and modifed graph：
graph = gs.import_onnx(onnx.load(ori_onnx_path))

I got the following error:
File "onnx_surgery.py", line 72, in onnx_model_surgery
graph = gs.import_onnx(onnx.load(ori_onnx_path))
File "/usr/local/lib/python3.8/dist-packages/onnx/init.py", line 134, in load_model
model = load_model_from_string(s, format=format)
File "/usr/local/lib/python3.8/dist-packages/onnx/init.py", line 171, in load_model_from_string
return _deserialize(s, ModelProto())
File "/usr/local/lib/python3.8/dist-packages/onnx/init.py", line 108, in _deserialize
decoded = cast(Optional[int], proto.ParseFromString(s))
File "/usr/local/lib/python3.8/dist-packages/google/protobuf/message.py", line 202, in ParseFromString
return self.MergeFromString(serialized)
File "/usr/local/lib/python3.8/dist-packages/google/protobuf/internal/python_message.py", line 1128, in MergeFromString
if self._InternalParse(serialized, 0, length) != length:
File "/usr/local/lib/python3.8/dist-packages/google/protobuf/internal/python_message.py", line 1178, in InternalParse
raise message_mod.DecodeError('Field number 0 is illegal.')
google.protobuf.message.DecodeError: Field number 0 is illegal.

And I have found another issued got the similar problem , I have tried the solustion

    onnx.shape_inference.infer_shapes_path(ori_onnx_path, opt_onnx_path)
    graph = gs.import_onnx(onnx.load(opt_onnx_path))

but the graph loaded from opt_onnx_path has no nodes and inputs, I have no idea how to edit the grah？ Is there any solution for my Scenario？

[MsDidin]

Is there any body follow this issue？

[zerollzeng]

@pranavm-nvidia ^ ^

[pranavm-nvidia]

Have you tried the instructions here?

[MsDidin]

Have you tried the instructions here?

Thanks a lot, this link solved my problem!