What does passing a training example produce? If the model is converged you would expect the output to be very close to the target, if your model produces the exact same output for a training data point I would be very suspicious that something has gone wrong.
Source: Discussion on r/MachineLearning




