Pull Request Test Coverage Report for Build 7277606585

• 0 of 0 changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 0.0%

Totals
Change from base Build 7277589910:	0.0%
Covered Lines:	0
Relevant Lines:	3682

💛 - Coveralls

I'm getting an error in the trainer notebook if I run it now (in the cell where GNN is tested). I get the same error if I run it locally from this branch and the latest env, but not from my working repo/env.

UnboundLocalError Traceback (most recent call last)
Cell In[9], line 7
4 earlystop_maxgap = 0.1
5 min_epoch = 10
----> 7 trainer.train(
8 nepoch = epochs,
9 batch_size = batch_size,
10 earlystop_patience = earlystop_patience,
11 earlystop_maxgap = earlystop_maxgap,
12 min_epoch = min_epoch,
13 validate = True,
14 filename = os.path.join(output_path, f"gnn_{task}", "model.pth.tar"))
16 epoch = trainer.epoch_saved_model
17 print(f"Model saved at epoch {epoch}")
File /opt/conda/envs/deeprank2/lib/python3.10/site-packages/deeprank2/trainer.py:640, in Trainer.train(self, nepoch, batch_size, shuffle, earlystop_patience, earlystop_maxgap, min_epoch, validate, num_workers, best_model, filename)
638 # Now that the training loop is over, save the model
639 if filename:
--> 640 torch.save(checkpoint_model, filename)
641 self.opt_loaded_state_dict = checkpoint_model["optimizer_state"]
642 self.model_load_state_dict = checkpoint_model["model_state"]
UnboundLocalError: local variable 'checkpoint_model' referenced before assignment

The issue is that if the loss is nan (such as in this case using very few datapoints), checkpoint_model is never created, thus the error. It has been fixed in #535