Building a PaddlePaddle Environment to Complete Text Sentiment Classification

Background

Based on Baidu's deep learning technology research and business application for many years, PaddlePaddle integrates with deep learning core training and reasoning framework, basic model base, end-to-end development kit, and abundant toolkits. This chapter takes bcc.vgn3 (Ubuntu18.04 LTS) as an example to introduce how to quickly build a GPU environment for PaddlePaddle and use the pre-trained model to complete text sentiment classification prediction.

Precondition

• Provide a GPU instance, such as gn3, gn2 and vgn3.
• Install Cuda and GPU driver in the GPU instance. It is recommended to install the Cuda version 11.2, 10.2, or 10.1.
• Purchase EIP to access the public network.

Operation Steps

Log in to the GPU instance, and view the Cuda and GPU driver versions.

 nvidia-smi

View the current Python environment and confirm that the python version is a3.6 / 3.7 / 3.8 / 3.9

 which python
 python --version

Install PIP3 and upgrade

 apt-get install python3-pip
 pip3 install -U pip

Install the paddlepaddle-gpu framework. This chapter takes Cuda 10.2 as an example. If you need to use other Cuda versions to build a PaddlePaddle environment, please see Installation of Other Cuda Versions.

 python -m pip install paddlepaddle-gpu -i https://mirror.baidu.com/pypi/simple

Install pre-trained model management tool paddlehub

 pip install paddlehub

Download the pre-trained model senta_bilstm by paddlehub

 hub install senta_bilstm==1.2.0

Run senta_bilstm

 hub run senta_bilstm --input_text “I love artificial intelligence"

The prediction results are as follows, and the model predicts that the text is positive emotion.