| 50 | |
| 51 | |
| 52 | def reload(): |
| 53 | print('This is reload') |
| 54 | # build entire net again and restore |
| 55 | tf_x = tf.placeholder(tf.float32, x.shape) # input x |
| 56 | tf_y = tf.placeholder(tf.float32, y.shape) # input y |
| 57 | l_ = tf.layers.dense(tf_x, 10, tf.nn.relu) # hidden layer |
| 58 | o_ = tf.layers.dense(l_, 1) # output layer |
| 59 | loss_ = tf.losses.mean_squared_error(tf_y, o_) # compute cost |
| 60 | |
| 61 | sess = tf.Session() |
| 62 | # don't need to initialize variables, just restoring trained variables |
| 63 | saver = tf.train.Saver() # define a saver for saving and restoring |
| 64 | saver.restore(sess, './params') |
| 65 | |
| 66 | # plotting |
| 67 | pred, l = sess.run([o_, loss_], {tf_x: x, tf_y: y}) |
| 68 | plt.subplot(122) |
| 69 | plt.scatter(x, y) |
| 70 | plt.plot(x, pred, 'r-', lw=5) |
| 71 | plt.text(-1, 1.2, 'Reload Loss=%.4f' % l, fontdict={'size': 15, 'color': 'red'}) |
| 72 | plt.show() |
| 73 | |
| 74 | |
| 75 | save() |