Fail to train a GCCN with a code example in the README.md file ?

[JieCoa]

When I clone the code to my local environment(Windows) and try to run the code example of "Defining 和 training a GCCN", I get some error and I don't know how to fix it out. Can anyone give me some advise?

[JieCoa]

And I also fail to reproduce experiments in "scripts/topotune/existing_models" with an error "In 'hydra/config': Could not find 'hydra/job_logging/colorlog'" (Is it a problem with my environment configuration?)

[gbg141]

Hi @JieCoa ! Thanks for reaching out and pointing this out. Actually, the scripts to reproduce TopoTune got outdated after the last changes that we applied to the library.

Our last PR (#124) addresses this, and now the scripts are compatible with the current code. For instance, you could run

python -m topobenchmark model=cell/topotune model.tune_gnn=GCN model.backbone.GNN.num_layers=1 model.backbone.neighborhoods=\[1-up_incidence-0,1-up_adjacency-1,1-down_incidence-2\] dataset=graph/MUTAG optimizer.parameters.lr=0.001 model.feature_encoder.out_channels=128 model.backbone.layers=2 model.readout.readout_name=PropagateSignalDown model.feature_encoder.proj_dropout=0.25 dataset.dataloader_params.batch_size=32 transforms.graph2cell_lifting.max_cell_length=10 dataset.split_params.data_seed=0,3,5,7,9 trainer.max_epochs=500 trainer.min_epochs=50 trainer.check_val_every_n_epoch=5 callbacks.early_stopping.patience=10

Please let us know if you still have some issues reproducing the experiments: probably then it would be related to an environment configuration, and we could take a closer look.

[JieCoa]

Thanks for your reply, but I still have some questions about the model.backbone.neighborhoods configuration. I used the script you provided with --multirun to test. And I got an unexpected error in the follow image.

Then I test the script in "Defining and training a GCCN".
python -m topobenchmark dataset=graph/PROTEINS dataset.split_params.data_seed=1 model=cell/topotune model.tune_gnn=GCN model.backbone.GNN.num_layers=2 model.backbone.neighborhoods=\[1-up_laplacian-0,1-down_incidence-2\] model.backbone.layers=4 model.feature_encoder.out_channels=32 model.feature_encoder.proj_dropout=0.3 model.readout.readout_name=PropagateSignalDown logger.wandb.project=TopoTune_cell trainer.max_epochs=1000 callbacks.early_stopping.patience=50
A different but somewhat similar error to the one mentioned above occurred.

I thought the error was related to the \ symbol in model.backbone.neighborhoods=\[1-up_laplacian-0,1-down_incidence-2\] , so I removed all the \ and then the code ran successfully.

python -m topobenchmark  dataset=graph/PROTEINS  dataset.split_params.data_seed=1  model=cell/topotune  model.tune_gnn=GCN  model.backbone.GNN.num_layers=2  model.backbone.neighborhoods=[1-up_laplacian-0,1-down_incidence-2]  model.backbone.layers=4  model.feature_encoder.out_channels=32  model.feature_encoder.proj_dropout=0.3  model.readout.readout_name=PropagateSignalDown  logger.wandb.project=TopoTune_cell  trainer.max_epochs=1000  callbacks.early_stopping.patience=50 

I don't know what the situation is like in your area, but this is my current situation. If I have misunderstood, I hope to get your reply.