Convert PLATO-2 model to ONNX format

[fadelma]

Hello everyone,

I'm currently trying to convert PLATO-2 model into ONNX format using Paddle2ONNX. However, when I try to convert the NSP model, I got this error:

Traceback (most recent call last):
  File "/usr/local/bin/paddle2onnx", line 8, in <module>
    sys.exit(main())
  File "/usr/local/lib/python3.7/dist-packages/paddle2onnx/command.py", line 184, in main
    input_shape_dict=input_shape_dict)
  File "/usr/local/lib/python3.7/dist-packages/paddle2onnx/command.py", line 148, in program2onnx
    operator_export_type=operator_export_type)
  File "/usr/local/lib/python3.7/dist-packages/paddle2onnx/convert.py", line 84, in program2onnx
    enable_onnx_checker, operator_export_type)
  File "/usr/local/lib/python3.7/dist-packages/paddle2onnx/convert.py", line 34, in export_onnx
    operator_export_type, verbose)
  File "/usr/local/lib/python3.7/dist-packages/paddle2onnx/graph/onnx_graph.py", line 240, in build
    onnx_graph = ONNXGraph(paddle_graph, opset_version=opset_version, operator_export_type=operator_export_type)
  File "/usr/local/lib/python3.7/dist-packages/paddle2onnx/graph/onnx_graph.py", line 79, in __init__
    self.update_opset_version()
  File "/usr/local/lib/python3.7/dist-packages/paddle2onnx/graph/onnx_graph.py", line 194, in update_opset_version
    self.opset_version = OpMapper.get_recommend_opset_version(node_map, self.opset_version)
  File "/usr/local/lib/python3.7/dist-packages/paddle2onnx/op_mapper/op_mapper.py", line 129, in get_recommend_opset_version
    node_map, opset_version, True)
  File "/usr/local/lib/python3.7/dist-packages/paddle2onnx/op_mapper/op_mapper.py", line 174, in check_support_status
    raise NotImplementedError(error_info)
NotImplementedError: 
There's 1 ops are not supported yet
=========== gather_nd ===========

Is this Paddle2ONNX's issue?

Also, Is there anyone who had successfully converted PLATO-2 model to ONNX format with Paddle2ONNX or other alternative methods and does not mind to share how to do it?

Thank you very much in advance!

[yeliang2258]

Hello, We have received your question, and we will solve it as soon as possible.

[yeliang2258]

Hi, please pull this PR for conversion test, thank you. https://github.com/PaddlePaddle/Paddle2ONNX/pull/479

[fadelma]

Hello @yeliang2258, I got this error when trying to convert the NSP model of PLATO-2 32L

Hello, I got this error when trying to convert the NSP model of PLATO-2 32L

2022-01-11 04:33:16 [INFO]      ONNX model generated is valid.
Traceback (most recent call last):
  File "/mnt/data3/shared_envs/paddle_env/lib/python3.9/site-packages/paddle2onnx-0.9.0-py3.9.egg/paddle2onnx/utils.py", line 43, in check_model
  File "/mnt/data3/shared_envs/paddle_env/lib/python3.9/site-packages/onnx/checker.py", line 101, in check_model
    protobuf_string = model.SerializeToString()
ValueError: Message onnx.ModelProto exceeds maximum protobuf size of 2GB: 6530750548

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/mnt/data3/shared_envs/paddle_env/bin/paddle2onnx", line 33, in <module>
    sys.exit(load_entry_point('paddle2onnx==0.9.0', 'console_scripts', 'paddle2onnx')())
  File "/mnt/data3/shared_envs/paddle_env/lib/python3.9/site-packages/paddle2onnx-0.9.0-py3.9.egg/paddle2onnx/command.py", line 187, in main
  File "/mnt/data3/shared_envs/paddle_env/lib/python3.9/site-packages/paddle2onnx-0.9.0-py3.9.egg/paddle2onnx/command.py", line 151, in program2onnx
  File "/mnt/data3/shared_envs/paddle_env/lib/python3.9/site-packages/paddle2onnx-0.9.0-py3.9.egg/paddle2onnx/convert.py", line 83, in program2onnx
  File "/mnt/data3/shared_envs/paddle_env/lib/python3.9/site-packages/paddle2onnx-0.9.0-py3.9.egg/paddle2onnx/convert.py", line 37, in export_onnx
  File "/mnt/data3/shared_envs/paddle_env/lib/python3.9/site-packages/paddle2onnx-0.9.0-py3.9.egg/paddle2onnx/graph/onnx_graph.py", line 234, in export_proto
  File "/mnt/data3/shared_envs/paddle_env/lib/python3.9/site-packages/paddle2onnx-0.9.0-py3.9.egg/paddle2onnx/utils.py", line 45, in check_model
Exception: ONNX model is not valid.

[jiangjiajun]

@fadelma The onnx checker will be failed while model_size > 2GB, we haven't test conversion on such large model, please try to disable onnx checker while converting paddle to onnx.

[fadelma]

Hello @jiangjiajun and @yeliang2258. I have disabled the onnx checker, but then I got this error

Traceback (most recent call last):
  File "/mnt/data3/shared_envs/paddle_env/bin/paddle2onnx", line 33, in <module>
    sys.exit(load_entry_point('paddle2onnx==0.9.0', 'console_scripts', 'paddle2onnx')())
  File "/mnt/data3/shared_envs/paddle_env/lib/python3.9/site-packages/paddle2onnx-0.9.0-py3.9.egg/paddle2onnx/command.py", line 187, in main
  File "/mnt/data3/shared_envs/paddle_env/lib/python3.9/site-packages/paddle2onnx-0.9.0-py3.9.egg/paddle2onnx/command.py", line 151, in program2onnx
  File "/mnt/data3/shared_envs/paddle_env/lib/python3.9/site-packages/paddle2onnx-0.9.0-py3.9.egg/paddle2onnx/convert.py", line 83, in program2onnx
  File "/mnt/data3/shared_envs/paddle_env/lib/python3.9/site-packages/paddle2onnx-0.9.0-py3.9.egg/paddle2onnx/convert.py", line 46, in export_onnx
ValueError: Message onnx.ModelProto exceeds maximum protobuf size of 2GB: 6530750548

[jiangjiajun]

Please send a same issue to Paddle2ONNX, this will take some effort to solve