ebm_sandbox.py [448:471]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    l1_norm = tf.placeholder(shape=(), dtype=tf.float32)
    l2_norm = tf.placeholder(shape=(), dtype=tf.float32)

    def compute_logit(X, stop_grad=False, num_steps=0):
        batch_size = tf.shape(X)[0]
        X = tf.reshape(X, (batch_size, 1, 32, 32, 3))
        X = tf.reshape(tf.tile(X, (1, 10, 1, 1, 1)), (batch_size * 10, 32, 32, 3))
        Y_new = tf.reshape(Y, (batch_size*10, 10))

        X_min = X - 8 / 255.
        X_max = X + 8 / 255.

        for i in range(num_steps):
            X = X + tf.random_normal(tf.shape(X), mean=0.0, stddev=0.005)

            energy_noise = model.forward(X, weights, label=Y, reuse=True)
            x_grad = tf.gradients(energy_noise, [X])[0]


            if FLAGS.proj_norm != 0.0:
                x_grad = tf.clip_by_value(x_grad, -FLAGS.proj_norm, FLAGS.proj_norm)

            X = X - FLAGS.step_lr * x_grad
            X = tf.maximum(tf.minimum(X, X_max), X_min)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



ebm_sandbox.py [675:698]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    l1_norm = tf.placeholder(shape=(), dtype=tf.float32)
    l2_norm = tf.placeholder(shape=(), dtype=tf.float32)

    def compute_logit(X, stop_grad=False, num_steps=0):
        batch_size = tf.shape(X)[0]
        X = tf.reshape(X, (batch_size, 1, 32, 32, 3))
        X = tf.reshape(tf.tile(X, (1, 10, 1, 1, 1)), (batch_size * 10, 32, 32, 3))
        Y_new = tf.reshape(Y, (batch_size*10, 10))

        X_min = X - 8 / 255.
        X_max = X + 8 / 255.

        for i in range(num_steps):
            X = X + tf.random_normal(tf.shape(X), mean=0.0, stddev=0.005)

            energy_noise = model.forward(X, weights, label=Y, reuse=True)
            x_grad = tf.gradients(energy_noise, [X])[0]


            if FLAGS.proj_norm != 0.0:
                x_grad = tf.clip_by_value(x_grad, -FLAGS.proj_norm, FLAGS.proj_norm)

            X = X - FLAGS.step_lr * x_grad
            X = tf.maximum(tf.minimum(X, X_max), X_min)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -