provide the detailed training code

proforma code below$-$

def train$\_$classifier$($args$,$ train, dev$)$:

# raise Exception$("$Not fully implemented yet"$)$

# The following code DOES NOT WORK but can be a starting point for your implementation

# Some suggested snippets to use:

model $=$ Transformer$(...)$

model.zero$\_$grad$()$

model.train$()$

optimizer $=$ optim.Adam$($model$.$parameters$(),$ lr$=1$e$-4)$

num$\_$epochs $=10$

for t in range$(0,$ num$\_$epochs$)$:

loss$\_$this$\_$epoch $=0\mathrm{.}0$

random.seed$($t$)$

# You can use batching if you'd like

ex$\_$idxs $=[$i for i in range$(0,$ len$($train$))]$

random.shuffle$($ex$\_$idxs$)$

loss$\_$fcn $=$ nn$.$NLLLoss$()$

for ex$\_$idx in ex$\_$idxs:

loss $=$ loss$\_$fcn$(...)$ # TODO: Run forward and compute loss

# model.zero$\_$grad$()$

# loss.backward$()$

# optimizer.step$()$

loss$\_$this$\_$epoch $+=$ loss.item$()$

model.eval$()$

return model