def svm$\_$train$($X$,$y$,$num$\_$iters,batch$\_$size,reg$\_$para,learning$\_$rate$)$:

# you are going to create the training loop and record the training loss each iteration.

W$,$num$\_$train,num$\_$classes $=$ svm$\_$init$($X$,$ y$)$ # init the model

loss$\_$history $=[]$ # this is a list to record all the loss during training

for iteration in range$($num$\_$iters$)$: # create this "for" loop to train the model.

# randomly sample the indexs of training data and find the corresponding data and label

# $\%\%\%\%\%\%\%\%\%\%\%\%\%\%$ implement your code below $(3$ lines$)\%\%\%\%\%\%\%\%\%\%\%\%\%\%$

sample$\_$idxs $=$ # generate the random index from the training dataset wite the same size of batchsize

X$\_$batch $=$ #find all input data X$\_$batch

y$\_$batch $=$ #find all corresponding label y$\_$batch

# $\%\%\%\%\%\%\%\%\%\%\%\%\%\%$ your code ends here $\%\%\%\%\%\%\%\%\%\%\%\%\%\%$

loss, grad $=$ svm$\_$loss$($X$\_$batch,y$\_$batch,W$,$reg$\_$para$)$ # we will implement this loss function in the following section

loss$\_$history.append$($loss$)$

W $=$ svm$\_$weight$\_$update$($W$,$grad,learning$\_$rate$)$ # we will implement this loss function in the following section

if iteration $\%100==0$:

print$(\text{'}$iteration $\%$d $/\%$d: loss $\%$f$\text{'}\%($iteration$,$ num$\_$iters, loss$))$

return loss$\_$history,W