Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
You can not select more than 25 topics
Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
Hongyi Zhang
925b86798b
|
1 year ago | |
|---|---|---|
| .. | ||
| .gitignore | 1 year ago | |
| README.md | 1 year ago | |
| entity_classify_mb_icdm.py | 1 year ago | |
| format_dgl.py | 1 year ago | |
| icdm2022_dataset.py | 1 year ago | |
| model.py | 1 year ago | |
README.md
DGL example
- Process session1 dataset.
mkdir -p /dataset/dgl_data/
python format_dgl.py --graph=/dataset/icdm2022_session1_edges.csv --node=/dataset/icdm2022_session1_nodes.csv --storefile=/dataset/dgl_data/icdm2022_session1
mkdir -p /dataset/dgl_data/icdm2022_session1
mv /dataset/dgl_data/icdm2022_session1.* /dataset/dgl_data/icdm2022_session1
cp /dataset/icdm2022_session1_train_labels.csv /dataset/dgl_data/icdm2022_session1/icdm2022_session1_labels.csv
cp /dataset/icdm2022_session1_test_ids.txt /dataset/dgl_data/icdm2022_session1/icdm2022_session1_test_ids.csv
- Process session2 dataset.
mkdir -p /dataset/dgl_data/
python format_dgl.py --graph=/dataset/icdm2022_session2/icdm2022_session2_edges.csv --node=/dataset/icdm2022_session2/icdm2022_session2_nodes.csv --storefile=/dataset/dgl_data/icdm2022_session2
mkdir -p /dataset/dgl_data/icdm2022_session2
mv /dataset/dgl_data/icdm2022_session2.* /dataset/dgl_data/icdm2022_session2
cp /dataset/icdm2022_session2/icdm2022_session2_test_ids.txt /dataset/dgl_data/icdm2022_session2/icdm2022_session2_test_ids.csv
- train on session1 dataset.
python entity_classify_mb_icdm.py --data-cpu --gpu 0 --n-hidden 256 --batch-size 2048 --n-bases 8 --fanout 150 --n-layers 2 --n-epoch 500 --dropout 0.3 --model-path ./session1_rgcn.pt --session session1 --lr 0.0002 --result-path result_session1_dgl.json --data-dir /dataset/dgl_data
- inference on session2 dataset with loaded model (trained from session1).
python entity_classify_mb_icdm.py --data-cpu --gpu 0 --n-hidden 256 --batch-size 2048 --n-bases 8 --fanout 150 --n-layers 2 --model-path ./session1_rgcn.pt --session session2 --load-model --result-path result_session2_dgl.json --data-dir /dataset/dgl_data