English | 简体中文

Model Export

After model training, we can export the inference model and deploy it using inference library, which achieves faster inference speed.

This tutorial will show how to export a trained model。

Acquire trained weight

After model training, the weight with the highest accuracy is saved in path/to/save/best_ model/model.pdparams.

For the convenience of this demo, we run the following commands to download the trained weight of PP-LiteSeg.

wget https://paddleseg.bj.bcebos.com/dygraph/cityscapes/pp_liteseg_stdc1_cityscapes_1024x512_scale0.5_160k/model.pdparams

Export the prediction Model

Run the following command in the root of PaddleSeg, the inference model is saved in output/inference_model.

python tools/export.py \
       --config configs/pp_liteseg/pp_liteseg_stdc1_cityscapes_1024x512_scale0.5_160k.yml \
       --model_path model.pdparams \
       --save_dir output/inference_model

Description of Exported Script Parameters

Parammeter	Purpose	Is Needed	Default
config	The path of config file	yes	-
model_path	The path of trained weight	no	-
save_dir	The save dir for the inference model	no	output/inference_model
input_shape	Set the input shape (N*C*H*W) of the inference model, such as --input_shape 1 3 1024 1024。if input_shape is not provided，the input shape of the inference model is [-1, 3, -1, -1]. If the image shape in prediction is fixed, you should set the input_shape.	no	None
output_op	Set the op that is appended to the inference model, should in [argmax, softmax, none]. PaddleSeg models outputs logits (N*C*H*W) by default. Adding argmax operation, we get the label for every pixel, the dimension of output is N*H*W. Adding softmax operation, we get the probability of different classes for every pixel.	no	argmax
with_softmax	Deprecated params, please use --output_op. Add softmax operator at the end of the network. Since PaddleSeg networking returns Logits by default, you can set it to True if you want the deployment model to get the probability value	no	False
without_argmax	Deprecated params, please use --output_op. Whether or not to add argmax operator at the end of the network. Since PaddleSeg networking returns Logits by default, we add argmax operator at the end of the network by default in order to directly obtain the prediction results for the deployment model	no	False

Note that:

• If you encounter shape-relevant issue, please try to set the input_shape.

Prediction Model Files

output/inference_model
  ├── deploy.yaml            # Config file of deployment
  ├── model.pdiparams        # Paramters of static model
  ├── model.pdiparams.info   # Additional information witch is not concerned generally
  └── model.pdmodel          # Static model file

After exporting prediction model, it can be deployed by the following methods.

Deployment scenarios	Inference library	Tutorial
Server (Nvidia GPU and X86 CPU) Python deployment	Paddle Inference	doc
Server (Nvidia GPU and X86 CPU) C++ deployment	Paddle Inference	doc
Mobile deployment	Paddle Lite	doc
Service-oriented deployment	Paddle Serving	doc
Web deployment	Paddle JS	doc