8df6b252447e160c2433f2f5e0f5343f7451876b,models/utils.py,,fc_layer,#,220

Before Change

        wd: weight decay
    Returns the fc layer
    W = weight("W", shape)
    b = bias("b", [shape[1]])
    // Add weight decay to W
    weight_decay = tf.multiply(tf.nn.l2_loss(W), wd, name="weight_loss")
    tf.add_to_collection(LOSSES_COLLECTION, weight_decay)
    return activation(tf.nn.bias_add(tf.matmul(input_x, W), b))


def batch_norm(layer_output, is_training_, decay=0.9):

After Change

    W = weight("W", shape, wd=wd)
    result = tf.matmul(input_x, W)
    if bias_term:
        b = bias("b", [shape[1]])
        result = tf.nn.bias_add(result, b)

    return activation(result)


def batch_norm(layer_output, is_training_, decay=0.9):

Instances