def train(self):
objs = dutils.AverageMeter()
top1 = dutils.AverageMeter()
top5 = dutils.AverageMeter()
self.model.train()
for step, (input, target) in enumerate(self.train_queue):
input = input.cuda(self.device, non_blocking=True)
target = target.cuda(self.device, non_blocking=True)
self.optimizer.zero_grad()
logits, logits_aux = self.model(input)
loss = self.criterion(logits, target)
if self.args.auxiliary:
loss_aux = self.criterion(logits_aux, target)
loss += self.args.auxiliary_weight*loss_aux
loss.backward()
nn.utils.clip_grad_norm_(self.model.parameters(), self.args.grad_clip)
self.optimizer.step()
prec1, prec5 = dutils.accuracy(logits, target, topk=(1,5))
n = input.size(0)
objs.update(loss.item(), n)
top1.update(prec1.item(), n)
top5.update(prec5.item(), n)
if step % args.report_freq == 0:
self.logger.info('train %03d %e %f %f', step, objs.avg, top1.avg, top5.avg)
return top1.avg, objs.avg
def train(train_loader, model, criterion, optimizer, epoch):
batch_time = dutils.AverageMeter()
data_time = dutils.AverageMeter()
losses = dutils.AverageMeter()
top1 = dutils.AverageMeter()
top5 = dutils.AverageMeter()
# switch to train mode
model.train()
end = time.time()
for i, (input, target) in enumerate(train_loader):
# measure data loading time
data_time.update(time.time() - end)
if args.gpu is not None:
input = input.cuda(args.gpu, non_blocking=True)
# Host to GPU copies are much faster when they originate from pinned (page-locked) memory.
# once you pin a tensor or storage, you can use asynchronous GPU copies
target = target.cuda(args.gpu, non_blocking=True)
# compute output
output = model(input)
loss = criterion(output, target)
# measure accuracy and record loss
prec1, prec5 = dutils.accuracy(output, target, topk=(1, 5))
losses.update(loss.item(), input.size(0))
top1.update(prec1[0], input.size(0))
top5.update(prec5[0], input.size(0))
# compute gradient and do SGD step
optimizer.zero_grad()
loss.backward()
optimizer.step()
if i % args.print_freq == 0:
print('Epoch: [{0}][{1}/{2}]\t'
'Time {batch_time.val:.3f} ({batch_time.avg:.3f})\t'
'Data {data_time.val:.3f} ({data_time.avg:.3f})\t'
'Loss {loss.val:.4f} ({loss.avg:.4f})\t'
'Prec@1 {top1.val:.3f} ({top1.avg:.3f})\t'
'Prec@5 {top5.val:.3f} ({top5.avg:.3f})'.format(
epoch,
i,
len(train_loader),
batch_time=batch_time,
data_time=data_time,
loss=losses,
top1=top1,
top5=top5))
def train(self):
objs = dutils.AverageMeter()
top1 = dutils.AverageMeter()
for step, (input, target) in enumerate(self.train_queue):
self.model.train()
n = input.size(0)
input = input.cuda(non_blocking=True)
target = target.cuda(non_blocking=True)
# Get a random minibatch from the search queue(validation set) with replacement
input_search, target_search = next(iter(self.valid_queue))
input_search = input_search.cuda(non_blocking=True)
target_search = target_search.cuda(non_blocking=True)
# Update the architecture parameters
self.architect.step(input,
target,
input_search,
target_search,
self.lr,
self.optimizer,
unrolled=self.args.sec_approx)
self.optimizer.zero_grad()
logits = self.model(input)
loss = self.criterion(logits, target)
loss.backward()
nn.utils.clip_grad_norm_(self.model.parameters(),
self.args.grad_clip)
# Update the network parameters
self.optimizer.step()
prec1 = dutils.accuracy(logits, target, topk=(1, ))[0]
objs.update(loss.item(), n)
top1.update(prec1.item(), n)
if step % args.report_freq == 0:
self.logger.info('model size: %f',
dutils.calc_parameters_count(self.model))
self.logger.info('train %03d loss: %e top1: %f', step,
objs.avg, top1.avg)
return top1.avg, objs.avg
def validate(val_loader, model, criterion):
batch_time = dutils.AverageMeter()
losses = dutils.AverageMeter()
top1 = dutils.AverageMeter()
top5 = dutils.AverageMeter()
# switch to evaluate mode
model.eval()
with torch.no_grad():
end = time.time()
for i, (input, target) in enumerate(val_loader):
if args.gpu is not None:
input = input.cuda(args.gpu, non_blocking=True)
target = target.cuda(args.gpu, non_blocking=True)
# compute output
output = model(input)
loss = criterion(output, target)
# measure accuracy and record loss
prec1, prec5 = dutils.accuracy(output, target, topk=(1, 5))
losses.update(loss.item(), input.size(0))
top1.update(prec1[0], input.size(0))
top5.update(prec5[0], input.size(0))
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
if i % args.print_freq == 0:
print('Test: [{0}/{1}]\t'
'Time {batch_time.val:.3f} ({batch_time.avg:.3f})\t'
'Loss {loss.val:.4f} ({loss.avg:.4f})\t'
'Prec@1 {top1.val:.3f} ({top1.avg:.3f})\t'
'Prec@5 {top5.val:.3f} ({top5.avg:.3f})'.format(
i,
len(val_loader),
batch_time=batch_time,
loss=losses,
top1=top1,
top5=top5))
print(' * Prec@1 {top1.avg:.3f} Prec@5 {top5.avg:.3f}'.format(
top1=top1, top5=top5))
def infer(self):
objs = dutils.AverageMeter()
top1 = dutils.AverageMeter()
top5 = dutils.AverageMeter()
self.model.eval()
with torch.no_grad():
for step, (input, target) in enumerate(self.valid_queue):
input = input.cuda(self.device, non_blocking=True)
target = target.cuda(self.device, non_blocking=True)
logits, _ = self.model(input)
loss = self.criterion(logits, target)
prec1, prec5 = dutils.accuracy(logits, target, topk=(1,5))
n = input.size(0)
objs.update(loss.item(), n)
top1.update(prec1.item(), n)
top5.update(prec5.item(), n)
if step % args.report_freq == 0:
self.logger.info('valid %03d %e %f %f', step, objs.avg, top1.avg, top5.avg)
return top1.avg, top5.avg, objs.avg
