目录

Jittor_warmup

This repository contains the solution files for the Jittor warmup task on Cora node classification.

Task

Train a GCN model with Jittor and JittorGeometric on the Cora citation graph, then predict the labels of test nodes. The required submission files are:

  • gcn.py: model definition, training, validation, and prediction script
  • result.json: predicted labels for test nodes

Files

.
|-- gcn.py
|-- result.json
|-- README.md
|-- requirements.txt
`-- .gitignore

Usage

Place the official contest data at data/cora.pkl, then run:

python gcn.py --data data/cora.pkl --output result.json

The script reads the pickle data, trains several GCN runs, selects the best validation run, and writes test-node predictions in the required JSON format.

Result

result.json contains 1000 predictions for node ids 1708 to 2707, with class ids from 0 to 6.

Notes

The official data/cora.pkl file is not included in this repository. It should be provided by the contest package.

关于
34.0 KB
邀请码
    Gitlink(确实开源)
  • 加入我们
  • 官网邮箱:gitlink@ccf.org.cn
  • QQ群
  • QQ群
  • 公众号
  • 公众号

版权所有:中国计算机学会技术支持:开源发展技术委员会
京ICP备13000930号-9 京公网安备 11010802047560号