How To Calculate Loss Without Updating Model In Tensorflow

To my understanding, below code will calculate loss and update parameters in the model at the same time. _, c = sess.run([optimizer, loss], feed_dict={x:x, y:y}) so how to calcula

Solution 1:

To calculate the loss without updating the model, simply run the loss operation, without the optimizer operation.

c = sess.run(loss, feed_dict={x:x, y:y})

Note that when you run sess.run([optimizer, loss], feed_dict={x:x, y:y}) you get the loss value before applying the updates, so running:

_, c1 = sess.run([optimizer, loss], feed_dict={x:x, y:y})
c2 = sess.run(loss, feed_dict={x:x, y:y})

Will still yield different values of c1 and c2, since c2 is the loss value after updating the model.