No change in training loss during Skorch training

I am using GridSearchCV in Skorch but during some of the loops there is no change in trainloss/valacc while for some there is. I am using very rudimentary parameters so the results should be fairly similar. For e.g in this sample nothing is happening and goes on exactly the same till the last epoch

  epoch    train_loss    valid_acc    valid_loss     dur
-------  ------------  -----------  ------------  ------
      1        0.6931       0.4961        0.6931  0.0885
      2        0.6931       0.4961        0.6931  0.0881
      3        0.6931       0.4961        0.6931  0.0921
      4        0.6931       0.4961        0.6931  0.0815
      5        0.6931       0.4961        0.6931  0.0790
      6        0.6931       0.4961        0.6931  0.0685
      7        0.6931       0.4961        0.6931  0.0853
      8        0.6931       0.4961        0.6931  0.0951
      9        0.6931       0.4961        0.6931  0.0873
     10        0.6931       0.4961        0.6931  0.1011
     11        0.6931       0.4961        0.6931  0.0944
     12        0.6931       0.4961        0.6931  0.0839
     13        0.6931       0.4961        0.6931  0.0758
     14        0.6931       0.4961        0.6931  0.0754
     15        0.6931       0.4961        0.6931  0.0832

But in the next iteration, things work fine and the trainloss/validloss improves

 epoch    train_loss    valid_acc    valid_loss     dur
-------  ------------  -----------  ------------  ------
      1        0.6931       0.4992        0.6931  0.0859
      2        0.6930       0.5429        0.6921  0.0925
      3        0.6907       0.5491        0.6840  0.0861
      4        0.6808       0.6147        0.6689  0.0797
      5        0.6633       0.6318        0.6484  0.0734
      6        0.6413       0.6989        0.6280  0.0757
      7        0.6224       0.7114        0.6168  0.0964
      8        0.6105       0.7254        0.6153  0.0949
      9        0.6068       0.7379        0.6112  0.0836
     10        0.5935       0.7426        0.6048  0.1007
     11        0.5811       0.7441        0.6021  0.0835
     12        0.5693       0.7644        0.5954  0.0776
     13        0.5612       0.7488        0.6001  0.0773
     14        0.5548       0.7878        0.5763  0.0806
     15        0.5338       0.7691        0.5811  0.0750

Does anyone know the reason behind this ?
(Also is there a way to stop fit(x,y) from printing all the loops ?)

ann = NeuralNetBinaryClassifier(module=Net66,max_epochs=20, lr = 0.01, optimizer=torch.optim.Adam,
criterion=nn.BCELoss,batch_size=800, iterator_train__shuffle=True)

params = {
    'lr': [0.001, 0.002],
    'max_epochs': [25, 50],
}

gs = GridSearchCV(ann, params, refit=False, scoring='accuracy', cv =2)

gs.fit(X, y)