Choose the false statements about gradient descent.$($more than one are wrong$)$

$1$ If the learning rate is set to be large, it can lead to divergence.

$2$ During training, the weight is adjusted in the direction of positive gradient.

$3$Training can can be done very fast if the learning rate is set to be small.

$4$ The learning rate is bigger than zero.

$5$ Loss function shifts around the minimum if the learning rate is set to be large.

$6$ The gradient is assessed again for the new weight vector in each step.

$7$ The weight is updated gradually in the direction of the negative gradient.

$8$ Slow convergence is caused by small learning rate.