Skip to content

Latest commit

 

History

History
executable file
·
86 lines (56 loc) · 6.76 KB

README_en.md

File metadata and controls

executable file
·
86 lines (56 loc) · 6.76 KB
Raw

Paddle2ONNX

简体中文 | English

1 Introduction

Paddle2ONNX supports the conversion of PaddlePaddle model format to ONNX model format. Through ONNX, it is possible to deploy Paddle models to various inference engines, including TensorRT/OpenVINO/MNN/TNN/NCNN, as well as other inference engines or hardware that support the ONNX open-source format.

2 Paddle2ONNX Environment Dependencies

Paddle2ONNX itself does not depend on other components, but we recommend using Paddle2ONNX in the following environments:

  • PaddlePaddle == 2.6.0
  • onnxruntime >= 1.10.0

3 Install Paddle2ONNX

If you only want to install Paddle2ONNX without the need for secondary development, you can quickly install Paddle2ONNX by executing the following code.

pip install paddle2onnx

If you wish to conduct secondary development on Paddle2ONNX, please follow the GitHub Source Code Installation Method to compile Paddle2ONNX.

4 Quick Start Tutorial

4.1 Get the PaddlePaddle Deployment Model

When Paddle2ONNX exports the model, it needs to pass in the deployment model format, including two files

  • model_name.pdmodel: Indicates the model structure
  • model_name.pdiparams: Indicates model parameters

4.2 Adjusting Paddle Models

If adjustments to the input and output of Paddle models are needed, you can visit the Paddle related tools for tutorials.

4.3 Using Command Line to Convert PaddlePaddle Models

You can convert Paddle models to ONNX models using the command line with the following command:

paddle2onnx --model_dir saved_inference_model \
            --model_filename model.pdmodel \
            --params_filename model.pdiparams \
            --save_file model.onnx

The adjustable conversion parameters are listed in the following table:

Parameter Parameter Description
--model_dir Configure directory path containing Paddle models
--model_filename [Optional] Configure the filename to store the network structure under --model_dir
--params_filename [Optional] Configure the name of the file to store model parameters under --model_dir
--save_file Specify the converted model save directory path
--opset_version [Optional] Configure the OpSet version converted to ONNX, currently supports multiple versions such as 7~16, the default is 9
--enable_onnx_checker [Optional] Configure whether to check the correctness of the exported ONNX model, it is recommended to turn on this switch, the default is False
--enable_auto_update_opset [Optional] Whether to enable the opset version automatic upgrade function, when the lower version of the opset cannot be converted, automatically select the higher version of the opset for conversion, the default is True
--deploy_backend [Optional] Inference engine for quantitative model deployment, supports onnxruntime, tensorrt or others, when other is selected, all quantization information is stored in the max_range.txt file, the default is onnxruntime
--save_calibration_file [Optional] TensorRT 8.X version deploys the cache file that needs to be read to save the path of the quantitative model, the default is calibration.cache
--version [Optional] View paddle2onnx version
--external_filename [Optional] When the exported ONNX model is larger than 2G, you need to set the storage path of external data, the recommended setting is: external_data
--export_fp16_model [Optional] Whether to convert the exported ONNX model to FP16 format, and use ONNXRuntime-GPU to accelerate inference, the default is False
--custom_ops [Optional] Export Paddle OP as ONNX's Custom OP, for example: --custom_ops '{"paddle_op":"onnx_op"}, default is {}

4.4 Pruning ONNX

If you need to adjust ONNX models, please refer to ONNX related tools

4.5 Optimize ONNX

If you have optimization needs for the exported ONNX model, we recommend using onnx-simplifier. You can also optimize the model using the following command.

5 Code Contribution

A thriving ecosystem requires everyone's collaborative efforts. Developers can refer to the Paddle2ONNX Contribution Guide to contribute code to Paddle2ONNX.

6 License

Provided under the Apache-2.0 license.

7 Thank you for the donation

  • Thanks to the PaddlePaddle team for providing server support for the CI infrastructure of Paddle2ONNX.
  • Thanks to community users chenwhql, luotao1, goocody, jeff41404, jzhang553, ZhengBicheng for donating a total of 10,000 RMB to the Paddle2ONNX PMC on March 28, 2024, for the development of Paddle2ONNX.