Standalone inference

[dmsgnn]

description

The aim of this functionality is to implement, by args parameter, the possibility to save the model from main script avoiding the inference part. A new script (or a main implementation) will be then created for inference only purpose, where the model will be uploaded and only inference will be executed.

requirements

• [x] The user must be able to train the model and save it
• [x] The user must be able to make inference using the trained model

[dmsgnn]

training

the file main_pyg.py has been modified to automatically save the model at the end of the script.

the model can be trained launching the main file from command line as

python main_pyg.py --dataset ogbg-molhiv --gnn gin --epochs 1 --batch_size 1 

inference

once the model has been saved, it can be used to perform inference only. It can be done using the new added args parameter --inference, which should be set to true. Then the model will be automatically retrieved using the other parameters. an example of inference only is provided below.

python main_pyg.py --dataset ogbg-molhiv --gnn gin --epochs 1 --batch_size 1 --inference true