Video 78 - Evaluation our model's prediction on straight line.

[dsnal]

In above video, I noticed that model_2.train( ) is not included epoch loop or above. Before that model, model.train is included in every model epoch loop. Why is that not included?

[mrdbourke]

Hey @dsnal,

Good question!

model.train() is the default state of the model.

So it doesn't always have to be called but it should usually be called after calling model.eval().

For example (the code is just an example and not a full training/testing loop):

# Training loop
model.train() # this is the default
for x in training_data:
    model(x)

# Testing loop
model.eval()
for x in test_data:
    model(x)

So if you're training: call model.train().

And if you're testing/making predictions: call model.eval()

[Sarthak-Singh04]

I had a same problem thanks for answering this question