datasets

loss.py

@@ -37,7 +37,7 @@ def masked_spectral_angle(y_true, y_pred):
     true_masked = F.normalize(true_masked, p=2, dim=1)
     # print(pred_masked.sum(dim=1))
     # print((pred_masked * true_masked).sum(dim=1).shape)
-    return 1 -2 * torch.acos((pred_masked * true_masked).sum(dim=1)).mean() / np.pi
+    return 1 -2 * torch.acos((pred_masked * true_masked).sum(dim=1).mean()) / np.pi
 
 
 def masked_pearson_correlation_distance(y_true, y_pred, reduce='mean'):

main_custom.py

@@ -24,6 +24,7 @@ def train(model, data_train, epoch, optimizer, criterion_rt, criterion_intensity
     for param in model.parameters():
         param.requires_grad = True
     if forward == 'both':
+        i=0
         for seq, charge, rt, intensity in data_train:
             rt, intensity = rt.float(), intensity.float()
             if torch.cuda.is_available():
@@ -41,6 +42,7 @@ def train(model, data_train, epoch, optimizer, criterion_rt, criterion_intensity
             optimizer.zero_grad()
             loss.backward()
             optimizer.step()
+            print(i,'/',len(data_train))
 
         if wandb is not None:
             wdb.log({"train rt loss": losses_rt / len(data_train), "train int loss": losses_int / len(data_train),