How to run your code with link prediction task?

[GabbySuwichaya]

Hello, again... I would like to ask about how to configure your script for running with link prediction tasks...
It is just that at the moment, I am a little confused, and I hope you could help explain...?

• From my understanding, if I want to run your demo with link prediction, I should just set the custom key --custom-key EV13NSO8-ES-Link as in the following command line ...is that right?

python3 SuperGAT/main.py   --dataset-class Planetoid    --dataset-name Cora   --custom-key EV13NSO8-ES-Link     --num-gpus-total 2  

• However, I get the following error:

Traceback (most recent call last):
  File "SuperGAT/main.py", line 475, in <module>
    many_seeds_result = run_with_many_seeds(main_args, num_total_runs, gpu_id=alloc_gpu[0])
  File "SuperGAT/main.py", line 407, in run_with_many_seeds
    ret = run(_args, gpu_id=gpu_id, **kwargs)
  File "SuperGAT/main.py", line 321, in run
    run_link_prediction=(perf_task_for_val == "Link"))
  File "SuperGAT/main.py", line 202, in test_model
    val_or_test_edge_y = batch.val_edge_y if val_or_test == "val" else batch.test_edge_y
AttributeError: 'Data' object has no attribute 'val_edge_y'

Also, I have noticed that the option is_link_gnn=False .... What does this mean?

Here is the print out of the options:

        - att_lambda: 11.346574532931719
        - attention_type: prob_mask_only
        - batch_size: 128
        - checkpoint_dir: ../SuperGAT_checkpoints
        - custom_key: EV13NSO8-ES-Link
        - data_num_splits: 1
        - data_root: ~/graph-data
        - data_sampler: None
        - data_sampling_num_hops: None
        - data_sampling_size: None
        - dataset_class: Planetoid
        - dataset_name: Cora
        - dropout: 0.6
        - early_stop_patience: 485
        - early_stop_queue_length: 484
        - early_stop_threshold_loss: 0.009317052513589488
        - early_stop_threshold_perf: 0.0011587124922279313
        - edge_sampling_ratio: 0.8
        - epochs: 490
        - gpu_deny_list: [1, 2, 3]
        - heads: 8
        - is_cgat_full: False
        - is_cgat_ssnc: False
        - is_link_gnn: False
        - is_super_gat: True
        - l1_lambda: 0.0
        - l2_lambda: 0.008228864972965771
        - link_lambda: 0.0
        - loss: None
        - lr: 0.005
        - m: 
        - model_name: GAT
        - neg_sample_ratio: 0.5
        - num_gpus_to_use: 1
        - num_gpus_total: 2
        - num_hidden_features: 8
        - num_layers: 2
        - out_heads: 8
        - perf_task_for_val: Link
        - perf_type: accuracy
        - pool_name: None
        - pretraining_noise_ratio: 0.0
        - save_model: False
        - save_plot: True
        - scaling_factor: None
        - seed: 42
        - start_epoch: 0
        - super_gat_criterion: None
        - task_type: Link_Prediction
        - to_undirected: False
        - to_undirected_at_neg: False
        - total_pretraining_epoch: 0
        - use_bn: False
        - use_early_stop: True
        - use_pretraining: False
        - val_interval: 1
        - verbose: 0
        

[dongkwan-kim]

if I want to run your demo with link prediction, I should just set the custom key --custom-key EV13NSO8-ES-Link as in the following command line ...is that right?

Change Planetoid to LinkPlanetoid will solve the issue.
This class is basically the same as the original one, but with train/val/test split of edges.

Also, I have noticed that the option is_link_gnn=False .... What does this mean?

You can think is_link_gnn flag is deprecated.
This is for the model I experimented with before but did not put into the paper.

[GabbySuwichaya]

Thank you so much. :)