def infer(valid_loader, model, criterion):
"""
targert
valid loop and record loss ,accuracy
parameters
train_loader : training dataloader
model : training model
criterion : loss function
return
average top 1 accuracy
"""
total_loss = utils.AvgrageMeter()
total_top1 = utils.AvgrageMeter()
aux_nums = len(model.aux_ind)
total_loss_aux = [utils.AvgrageMeter() for i in range(aux_nums)]
total_top1_aux = [utils.AvgrageMeter() for i in range(aux_nums)]
model.eval()
for step, (data, target) in enumerate(valid_loader):
data = data.cuda()
target = target.cuda()
with torch.no_grad():
logits, auxs = model(data)
loss = F.cross_entropy(logits, target).cuda()
loss_aux = [F.cross_entropy(i, target).cuda() for i in auxs]
prec1 = utils.accuracy(logits, target)[0]
prec1_aux = [utils.accuracy(i, target)[0] for i in auxs]
n = data.size(0)
total_loss.update(loss.item(), n)
total_top1.update(prec1.item(), n)
[
total_loss_aux[ind].update(i.item(), n)
for ind, i in enumerate(loss_aux)
]
[
total_top1_aux[ind].update(i.item(), n)
for ind, i in enumerate(prec1_aux)
]
if (step + 1) % args.report_freq == 0:
[
logging.info('aux_{} {:0>3d} {:.6f} {:.3f}'.format(
i, step, total_loss_aux[i].avg, total_top1_aux[i].avg))
for i in range(aux_nums)
]
logging.info('\nvalid {:0>3d} {:.6f} {:.3f}'.format(
step, total_loss.avg, total_top1.avg))
del loss, logits, loss_aux
return total_top1.avg
def validate(epoch, val_loader, model, criterion, args):
batch_time = utils.AverageMeter('Time', ':6.3f')
losses = utils.AverageMeter('Loss', ':.4e')
top1 = utils.AverageMeter('Acc@1', ':6.2f')
top5 = utils.AverageMeter('Acc@5', ':6.2f')
if args.use_dali:
num_iter = val_loader._size // args.batch_size
else:
num_iter = len(val_loader)
model.eval()
with torch.no_grad():
end = time.time()
if args.use_dali:
for batch_idx, batch_data in enumerate(val_loader):
images = batch_data[0]['data'].cuda()
targets = batch_data[0]['label'].squeeze().long().cuda()
# compute output
logits = model(images)
loss = criterion(logits, targets)
# measure accuracy and record loss
pred1, pred5 = utils.accuracy(logits, targets, topk=(1, 5))
n = images.size(0)
losses.update(loss.item(), n)
top1.update(pred1[0], n)
top5.update(pred5[0], n)
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
else:
for batch_idx, (images, targets) in enumerate(val_loader):
images = images.cuda()
targets = targets.cuda()
# compute output
logits = model(images)
loss = criterion(logits, targets)
# measure accuracy and record loss
pred1, pred5 = utils.accuracy(logits, targets, topk=(1, 5))
n = images.size(0)
losses.update(loss.item(), n)
top1.update(pred1[0], n)
top5.update(pred5[0], n)
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
logger.info(' * Acc@1 {top1.avg:.3f} Acc@5 {top5.avg:.3f}'.format(
top1=top1, top5=top5))
return losses.avg, top1.avg, top5.avg
def infer(valid_queue, model, criterion):
objs = utils.AverageMeter()
top1 = utils.AverageMeter()
top5 = utils.AverageMeter()
model.eval()
with torch.no_grad():
for step, (data, target) in enumerate(valid_queue):
data = data.cuda()
target = target.cuda()
logits, _ = model(data)
loss = criterion(logits, target)
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
n = data.size(0)
objs.update(loss.item(), n)
top1.update(prec1.item(), n)
top5.update(prec5.item(), n)
if step % args.report_freq == 0:
logging.info('valid %03d %e %f %f', step, objs.avg, top1.avg,
top5.avg)
return top1.avg, top5.avg, objs.avg
def infer(valid_queue, model, criterion):
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
top5 = utils.AvgrageMeter()
model.eval()
with torch.no_grad():
for step, (input, target) in enumerate(valid_queue):
input = input.cuda()
target = target.cuda(non_blocking=True)
logits = model(input)
loss = criterion(logits, target)
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
n = input.size(0)
objs.update(loss.data, n)
top1.update(prec1.data, n)
top5.update(prec5.data, n)
if step % args.report_freq == 0:
logging.info('valid %03d %e %f %f', step, objs.avg, top1.avg,
top5.avg)
if 'debug' in args.save:
break
return top1.avg, objs.avg
def train(train_queue, model, criterion, optimizer):
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
top5 = utils.AvgrageMeter()
model.train()
for step, (input, target) in enumerate(train_queue):
target = target.cuda(async=True)
input = input.cuda()
input = Variable(input)
target = Variable(target)
optimizer.zero_grad()
logits, logits_aux = model(input)
loss = criterion(logits, target)
if args.auxiliary:
loss_aux = criterion(logits_aux, target)
loss += args.auxiliary_weight * loss_aux
loss.backward()
nn.utils.clip_grad_norm(model.parameters(), args.grad_clip)
optimizer.step()
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
n = input.size(0)
objs.update(loss.data[0], n)
top1.update(prec1.data[0], n)
top5.update(prec5.data[0], n)
if step % args.report_freq == 0:
logging.info('train %03d %e %f %f', step, objs.avg, top1.avg, top5.avg)
return top1.avg, objs.avg
def train(train_queue, model, criterion, optimizer):
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
model.train()
for step, (input, target) in enumerate(train_queue):
input = input.cuda(non_blocking=True)
target = target.cuda(non_blocking=True)
optimizer.zero_grad()
logits, logits_aux = model(input)
loss = criterion(logits, target)
if args.auxiliary:
loss_aux = criterion(logits_aux, target)
loss += args.auxiliary_weight * loss_aux
loss.backward()
nn.utils.clip_grad_norm_(model.parameters(), args.grad_clip)
optimizer.step()
prec1, _ = utils.accuracy(logits, target, topk=(1, 5))
n = input.size(0)
objs.update(loss.data.item(), n)
top1.update(prec1.data.item(), n)
if step % args.report_freq == 0:
logging.info('Train Step: %03d Objs: %e Acc: %f', step, objs.avg,
top1.avg)
return top1.avg, objs.avg
def train(train_queue, model, optimizer, gpu):
objs = utils.AverageTracker()
top1 = utils.AverageTracker()
top5 = utils.AverageTracker()
model.train()
for step, (input, target) in enumerate(train_queue):
# input = input.cuda(non_blocking=True)
# target = target.cuda(non_blocking=True)
if gpu:
target = target.cuda(non_blocking=True)
optimizer.zero_grad()
dic = model(input, target)
logits = dic['logits']
loss = dic['loss']
loss.backward()
nn.utils.clip_grad_norm_(model.parameters(), args.grad_clip)
optimizer.step()
prec1, prec5 = utils.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:
logging.info('train %03d %e %f %f', step, objs.avg, top1.avg, top5.avg)
del loss, logits, prec1, prec5, input, target
return top1.avg, objs.avg
def infer(valid_queue, model, criterion):
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
top5 = utils.AvgrageMeter()
model.eval()
for step, (input, target) in enumerate(valid_queue):
input = Variable(input, volatile=True).cuda()
target = Variable(target, volatile=True).cuda(async=True)
logits = model(input)
loss = criterion(logits, target)
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
n = input.size(0)
objs.update(loss.data[0], n)
top1.update(prec1.data[0], n)
top5.update(prec5.data[0], n)
if step % args.report_freq == 0:
logging.info('valid %03d %e %f %f', step, objs.avg, top1.avg,
top5.avg)
if args.debug:
break
return top1.avg, objs.avg
def train(train_queue, model, criterion, optimizer):
objs = utils.AverageMeter()
top1 = utils.AverageMeter()
model.train()
for step, (inp, target) in enumerate(train_queue):
inp = inp.cuda(non_blocking=True)
target = target.cuda(non_blocking=True)
model.zero_grad()
logits = model(inp)
loss = criterion(logits, target)
loss.backward()
optimizer.step()
prec1, _ = utils.accuracy(logits, target, topk=(1, 5))
n = inp.size(0)
objs.update(loss.clone().item(), n)
top1.update(prec1.clone().item(), n)
step += 1
if step % args.report_freq == 0:
print('Train Step: {:3d} Objs: {:.4f} Acc: {:.2f}'.format(step, objs.avg, top1.avg))
return top1.avg, objs.avg
def infer(valid_queue, model, criterion):
objs = utils.AverageMeter()
top1 = utils.AverageMeter()
model.eval()
with torch.no_grad():
for step, (input, target) in enumerate(valid_queue):
input = Variable(input)
target = Variable(target)
if args.gpu != -1:
input = input.cuda(non_blocking=True)
target = target.cuda(non_blocking=True)
logits = model(input)
loss = criterion(logits, target)
prec1 = utils.accuracy(logits, target, topk=(1, ))
n = input.size(0)
objs.update(loss.detach().item(), n)
top1.update(prec1[0].item(), n)
if step % args.report_freq == 0:
logging.info('valid %03d %e %f', step, objs.avg, top1.avg)
return top1.avg, objs.avg
def test(test_queue, model, criterion, logger):
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
with torch.no_grad():
for step, (input, target) in enumerate(test_queue):
n = input.size(0)
input = Variable(input.float()).to(device)
target = Variable(target.long()).to(device)
logits = model(input)
loss = criterion(logits, target)
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
objs.update(loss.item(), n)
top1.update(prec1.item(), n)
if step % args.report_freq == 0:
logger.info('time = %s, test %03d %e %f',
str(utils.get_unix_timestamp()), step, objs.avg,
top1.avg)
print('time = {}, test {} {}'.format(
str(utils.get_unix_timestamp()), step, objs.avg))
return objs.avg, top1.avg
def train(train_queue, model, criterion, optimizer, logger):
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
for step, (input, target) in enumerate(train_queue):
n = input.size(0)
input = Variable(input.float(), requires_grad=False).to(device)
target = Variable(target.long(), requires_grad=False).to(device)
logits = model(input)
loss = criterion(logits, target)
optimizer.zero_grad()
loss.backward()
nn.utils.clip_grad_norm_(model.parameters(), args.grad_clip)
optimizer.step()
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
objs.update(loss.item(), n)
top1.update(prec1.item(), n)
if step % args.report_freq == 0:
logger.info('time = %s, train %03d %e %f',
str(utils.get_unix_timestamp()), step, objs.avg,
top1.avg)
print('time = {}, train {} {}'.format(
str(utils.get_unix_timestamp()), step, objs.avg))
return objs.avg, top1.avg
def infer(valid_queue, model, criterion, writer_dict):
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
top5 = utils.AvgrageMeter()
model.eval()
# tensorboard logger
writer = writer_dict['writer']
val_step = writer_dict['val_steps']
for step, (input, target) in enumerate(valid_queue):
input = Variable(input, volatile=True).cuda()
# target = Variable(target, volatile=True).cuda(async=True)
target = Variable(target, volatile=True).cuda()
logits = model(input)
loss = criterion(logits, target)
writer.add_scalar('val_loss', loss.data, val_step)
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
writer.add_scalar('val_prec1', prec1.data, val_step)
writer.add_scalar('val_prec5', prec5.data, val_step)
n = input.size(0)
objs.update(loss.data, n)
top1.update(prec1.data, n)
top5.update(prec5.data, n)
val_step += 1
writer_dict['val_steps'] += 1
if step % args.report_freq == 0:
logging.info('valid %03d %e %f %f', step, objs.avg, top1.avg,
top5.avg)
return top1.avg, objs.avg
def infer(valid_queue, model, criterion, temperature, logger=None, batch_num=-1,
log_frequence=10):
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
top5 = utils.AvgrageMeter()
model.eval()
# logger.info("Start new epoch inference")
tic = time.time()
for step, (input, target) in enumerate(valid_queue):
input = input.cuda()
target = target.cuda()
logits , _, _ = model(input , temperature)
loss = criterion(logits, target)
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
n = input.size(0)
objs.update(loss.data , n)
top1.update(prec1.data , n)
top5.update(prec5.data , n)
if logger is not None:
if step > 0 and step % log_frequence == 0:
toc = time.time()
speed = 1.0 * log_frequence * n / (toc - tic)
if batch_num > 0:
logger.info("Step[%d/%d] speed[%.4f samples/s] loss[%.6f] acc[%.4f]" % (step, batch_num,
speed, loss.detach().cpu().numpy(), \
prec1.detach().cpu().numpy() / 100.0))
else:
logger.info("Step[%d] speed[%.4f samples/s] loss[%.6f] acc[%.4f]" % (step,
speed, loss.detach().cpu().numpy(), \
prec1.detach().cpu().numpy() / 100.0))
tic = time.time()
return top1.avg, top5.avg, objs.avg
def infer(valid_queue, model, criterion, args, gpu):
'''
在最后一个epoch后打印验证集计算结果
'''
print('正在进行测试！！！！')
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
top5 = utils.AvgrageMeter()
model.eval()
for step, (input, target) in enumerate(valid_queue):
input = input.cuda(non_blocking=True)
target = target.cuda(non_blocking=True)
logits = model(input) # 计算预测结果
loss = criterion(logits, target)
prec1, prec5 = utils.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 gpu == 0:
if step % args.report_freq == 0:
logging.info('valid %03d %e %f %f', step, objs.avg, top1.avg, top5.avg)
print('valid %03d %e %f %f', step, objs.avg, top1.avg, top5.avg)
return top1.avg, objs.avg
def train(train_queue, model, criterion, optimizer):
global is_multi_gpu
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
top5 = utils.AvgrageMeter()
model.train()
for step, (input, target) in enumerate(train_queue):
n = input.size(0)
input = input.cuda()
target = target.cuda(non_blocking=True)
optimizer.zero_grad()
logits, logits_aux = model(input)
loss = criterion(logits, target)
if args.auxiliary:
loss_aux = criterion(logits_aux, target)
loss += args.auxiliary_weight * loss_aux
loss.backward()
parameters = model.module.parameters() if is_multi_gpu else model.parameters()
nn.utils.clip_grad_norm_(parameters, args.grad_clip)
optimizer.step()
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
objs.update(loss.item(), n)
top1.update(prec1.item(), n)
top5.update(prec5.item(), n)
if step % args.report_freq == 0:
logging.info('train %03d %e %f %f', step, objs.avg, top1.avg, top5.avg)
wandb.log({"evaluation_train_accuracy_avg": objs.avg}, step=step)
wandb.log({"evaluation_train_accuracy_top1": top1.avg}, step=step)
wandb.log({"evaluation_train_accuracy_top5": top5.avg}, step=step)
return top1.avg, objs.avg
def infer(valid_queue, model, criterion):
global is_multi_gpu
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
top5 = utils.AvgrageMeter()
# nn.BatchNorm layers will use their running stats (in the default mode) and nn.Dropout will be deactivated.
model.eval()
for step, (input, target) in enumerate(valid_queue):
with torch.no_grad():
input = input.cuda()
target = target.cuda(non_blocking=True)
logits, _ = model(input)
loss = criterion(logits, target)
prec1, prec5 = utils.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:
logging.info('valid %03d %e %f %f', step, objs.avg, top1.avg, top5.avg)
return top1.avg, objs.avg
def train(train_loader, model, controller, optimizer, args):
total_loss = search_policies.cnn.utils.AverageMeter()
total_top1 = search_policies.cnn.utils.AverageMeter()
for step, (data, target) in enumerate(train_loader):
if args.debug and step > 10:
logging.warning("Breaking after 10 steps")
break
model.train()
n = data.size(0)
data = data.cuda()
target = target.cuda()
optimizer.zero_grad()
controller.eval()
dag, _, _ = controller()
logits, _ = model(data, dag)
loss = F.cross_entropy(logits, target)
loss.backward()
optimizer.step()
prec1 = utils.accuracy(logits, target)[0]
total_loss.update(loss.item(), n)
total_top1.update(prec1.item(), n)
if step % args.report_freq == 0:
logging.info('train %03d %e %f', step, total_loss.avg,
total_top1.avg)
return total_top1.avg
def validate(epoch, val_loader, model, criterion, args):
batch_time = utils.AverageMeter('Time', ':6.3f')
losses = utils.AverageMeter('Loss', ':.4e')
top1 = utils.AverageMeter('Acc@1', ':6.2f')
top5 = utils.AverageMeter('Acc@5', ':6.2f')
# switch to evaluation mode
model.eval()
with torch.no_grad():
end = time.time()
for i, (images, target) in enumerate(val_loader):
images = images.cuda()
target = target.cuda()
# compute output
logits = model(images)
loss = criterion(logits, target)
# measure accuracy and record loss
pred1, pred5 = utils.accuracy(logits, target, topk=(1, 5))
n = images.size(0)
losses.update(loss.item(), n)
top1.update(pred1[0], n)
top5.update(pred5[0], n)
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
logger.info(' * Acc@1 {top1.avg:.3f} Acc@5 {top5.avg:.3f}'
.format(top1=top1, top5=top5))
return losses.avg, top1.avg, top5.avg
def child_test(self, test_queue, model, arch, criterion, verbose=True):
utils = self.utils
objs = search_policies.cnn.utils.AverageMeter()
top1 = search_policies.cnn.utils.AverageMeter()
top5 = search_policies.cnn.utils.AverageMeter()
model.eval()
# arch_l = arch
arch = self.process_arch(arch)
with torch.no_grad():
for step, (input, target) in enumerate(test_queue):
if self.args.debug:
if step > 10:
print("Break after 10 batch")
break
input = input.cuda()
target = target.cuda()
logits, _ = model(input, arch, bn_train=False)
loss = criterion(logits, target)
prec1, prec5 = utils.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 % self.args.report_freq == 0 and step > 0 and verbose:
logging.info(
'test | step %03d | loss %e | acc %f | acc-5 %f', step,
objs.avg, top1.avg, top5.avg)
return top1.avg, top5.avg, objs.avg
def infer(valid_queue, model, gpu):
objs = utils.AverageTracker()
top1 = utils.AverageTracker()
top5 = utils.AverageTracker()
model.eval()
with torch.no_grad():
for step, (input, target) in enumerate(valid_queue):
# input = input.cuda(non_blocking=True)
if gpu:
target = target.cuda(non_blocking=True)
dic = model(input, target)
logits = dic['logits']
loss = dic['loss']
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
n = input.size(0)
assert isinstance(n, int)
objs.update(loss.item(), n)
top1.update(prec1.item(), n)
top5.update(prec5.item(), n)
if step % args.report_freq == 0:
logging.info('valid %03d %e %f %f', step, objs.avg, top1.avg, top5.avg)
del loss, logits, prec1, prec5, input, target
return top1.avg, top5.avg, objs.avg
def train(train_queue, model, optimizer, global_step, criterion):
objs = search_policies.cnn.utils.AverageMeter()
top1 = search_policies.cnn.utils.AverageMeter()
top5 = search_policies.cnn.utils.AverageMeter()
model.train()
for step, (input, target) in enumerate(train_queue):
input = input.cuda().requires_grad_()
target = target.cuda()
optimizer.zero_grad()
logits, aux_logits = model(input, global_step)
global_step += 1
loss = criterion(logits, target)
if aux_logits is not None:
aux_loss = criterion(aux_logits, target)
loss += 0.4 * aux_loss
loss.backward()
nn.utils.clip_grad_norm_(model.parameters(), args.grad_bound)
optimizer.step()
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
n = input.size(0)
objs.update(loss.data, n)
top1.update(prec1.data, n)
top5.update(prec5.data, n)
if (step+1) % 100 == 0:
logging.info('train %03d loss %e top1 %f top5 %f', step+1, objs.avg, top1.avg, top5.avg)
return top1.avg, objs.avg, global_step
def infer(valid_loader, model, controller, criterion):
total_loss = utils.AvgrageMeter()
total_top1 = utils.AvgrageMeter()
model.eval()
controller.eval()
with torch.no_grad():
for step in range(20):
data, target = valid_loader.next_batch()
data = data.cuda()
target = target.cuda()
dag, _, _ = controller()
logits, auxs = model(dag, data)
loss = criterion(logits, target).cuda()
prec1 = utils.accuracy(logits, target)[0]
n = data.size(0)
total_loss.update(loss.item(), n)
total_top1.update(prec1.item(), n)
logging.info('valid {:0>3d} {:.6f} {:.3f}'.format(
step, loss.item(), prec1.item()))
logging.info('{}'.format([i for i in dag]))
with open(os.path.join(args.save, 'dag_all.txt'), 'a') as f:
f.write('{:.3f} {} infer\n'.format(prec1.item(),
[i for i in dag])) #
del loss, logits
return total_top1.avg
def infer(valid_queue, model, criterion):
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
top5 = utils.AvgrageMeter()
model.eval()
with torch.no_grad():
for step, (input, target) in enumerate(valid_queue):
input = input.cuda()
target = target.cuda(non_blocking=True)
logits, _ = model(input)
loss = criterion(logits, target)
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
n = input.size(0)
objs.update(loss.data, n)
top1.update(prec1.data, n)
top5.update(prec5.data, n)
if step % args.report_freq == 0:
end_time = time.time()
if step == 0:
duration = 0
start_time = time.time()
else:
duration = end_time - start_time
start_time = time.time()
logging.info(
'VALID Step: %03d Objs: %e R1: %f R5: %f Duration: %ds',
step, objs.avg, top1.avg, top5.avg, duration)
return top1.avg, top5.avg, objs.avg
def infer(valid_queue, model, epoch, criterion, writer, model_name):
batch_time = utils.AverageMeters('Time', ':6.3f')
losses = utils.AverageMeters('Loss', ':.4e')
top1 = utils.AverageMeters('Acc@1', ':6.2f')
top5 = utils.AverageMeters('Acc@5', ':6.2f')
model.eval()
progress = utils.ProgressMeter(len(valid_queue),
batch_time,
losses,
top1,
top5,
prefix='Test: ')
cur_step = epoch * len(valid_queue)
end = time.time()
for step, (input, target) in enumerate(valid_queue):
#input = input.cuda()
#target = target.cuda(non_blocking=True)
input = Variable(input, volatile=True).cuda()
#target = Variable(target, volatile=True).cuda(async=True)
target = Variable(target, volatile=True).cuda()
if model_name == "ShuffleNas":
block_choices = model.module.random_block_choices(
select_predefined_block=False)
ignore_first_two_cs = True # 0.2 and 0.4 scales are ignored
if config.cs_warm_up:
full_channel_mask, _ = model.module.random_channel_mask(
select_all_channels=config.use_all_channels,
epoch_after_cs=epoch - config.epoch_start_cs,
ignore_first_two_cs=ignore_first_two_cs)
else:
full_channel_mask, _ = model.module.random_channel_mask(
select_all_channels=config.use_all_channels)
logits = model(input, block_choices, full_channel_mask)
else:
logits = model(input, None, None)
loss = criterion(logits, target)
acc1, acc5 = utils.accuracy(logits, target, topk=(1, 5))
n = input.size(0)
reduced_loss = reduce_tensor(loss.data, world_size=config.world_size)
acc1 = reduce_tensor(acc1, world_size=config.world_size)
acc5 = reduce_tensor(acc5, world_size=config.world_size)
losses.update(to_python_float(reduced_loss), n)
top1.update(to_python_float(acc1), n)
top5.update(to_python_float(acc5), n)
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
if step % config.print_freq == 0:
progress.print(step)
logger.info('valid %03d %e %f %f', step, losses.avg, top1.avg,
top5.avg)
writer.add_scalar('val/loss', losses.avg, cur_step)
writer.add_scalar('val/top1', top1.avg, cur_step)
writer.add_scalar('val/top5', top5.avg, cur_step)
return top1.avg, losses.avg
def train(train_queue, model, criterion, optimizer):
objs = utils.AverageMeter()
top1 = utils.AverageMeter()
model.train()
for step, (input, target) in enumerate(train_queue):
input = input.cuda(non_blocking=True)
target = target.cuda(non_blocking=True)
model.zero_grad()
model.generate_share_alphas()
logits = model(input)
loss = criterion(logits, target)
loss.backward()
if args.grad_clip > 0:
nn.utils.clip_grad_norm_(model.parameters(), args.grad_clip)
optimizer.step()
prec1, _ = utils.accuracy(logits, target, topk=(1, 5))
n = input.size(0)
objs.update(loss.clone().item(), n)
top1.update(prec1.clone().item(), n)
if (step + 1) % args.report_freq == 0:
logging.info('Train Step: %03d Objs: %e Acc: %.2f', step + 1,
objs.avg, top1.avg)
return top1.avg, objs.avg
def infer(valid_queue, model, criterion):
global is_multi_gpu
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
top5 = utils.AvgrageMeter()
model.eval()
for step, (input, target) in enumerate(valid_queue):
input = input.cuda()
target = target.cuda()
logits = model(input)
loss = criterion(logits, target)
prec1, prec5 = utils.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:
logging.info('valid %03d %e %f %f', step, objs.avg, top1.avg,
top5.avg)
return top1.avg, objs.avg, loss
def infer(valid_queue, model, criterion):
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
top5 = utils.AvgrageMeter()
model.eval()
with torch.no_grad():
for step, (input, target) in enumerate(valid_queue):
# input = Variable(input, volatile=True).cuda()
# target = Variable(target, volatile=True).cuda(async=True)
input = input.to(device)
target = target.to(device)
logits, _ = model(input)
loss = criterion(logits, target)
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
n = input.size(0)
loss_data = loss.data[0] if loss.dim() != 0 else loss.item()
prec1_data = prec1.data[0] if prec1.dim() != 0 else prec1.item()
prec5_data = prec5.data[0] if prec5.dim() != 0 else prec5.item()
objs.update(loss_data, n)
top1.update(prec1_data, n)
top5.update(prec5_data, n)
if step % args.report_freq == 0:
logging.info('valid %03d %e %f %f', step, objs.avg, top1.avg,
top5.avg)
return top1.avg, objs.avg
def train(train_queue, model, criterion, optimizer):
objs = utils.AverageMeter()
top1 = utils.AverageMeter()
model.train()
for step, (inp, target) in enumerate(train_queue):
global global_step
global_step += 1
inp = inp.cuda(non_blocking=True)
target = target.cuda(non_blocking=True)
model.zero_grad()
logits = model(inp)
loss = criterion(logits, target)
loss.backward()
if args.grad_clip > 0:
nn.utils.clip_grad_norm_(model.parameters(), args.grad_clip)
optimizer.step()
prec1, _ = utils.accuracy(logits, target, topk=(1, 5))
n = inp.size(0)
objs.update(loss.clone().item(), n)
top1.update(prec1.clone().item(), n)
writer_tr.add_scalar('loss', loss.item(), global_step)
writer_tr.add_scalar('acc', prec1.item(), global_step)
decomp_loss_1 = torch.mean(torch.log(torch.sum(torch.exp(logits - torch.max(logits, 1)[0].view(-1, 1)), 1))).item()
writer_tr.add_scalar('decomp_loss_lgsmex', decomp_loss_1, global_step)
decomp_loss_2 = torch.mean(torch.max(logits, 1)[0] - logits[range(len(inp)), target.cpu().numpy()]).item()
writer_tr.add_scalar('decomp_loss_max', decomp_loss_2, global_step)
if (step + 1) % args.report_freq == 0:
logging.info('Train Step: %03d Objs: %e Acc: %.2f', step + 1, objs.avg, top1.avg)
return top1.avg, objs.avg
def infer(valid_queue, model, alpha, criterion):
"""Run model in eval only mode."""
objs = utils.AverageMeter()
top1 = utils.AverageMeter()
top5 = utils.AverageMeter()
# model.eval()
with torch.no_grad():
for step, (data, target) in enumerate(valid_queue):
n = data.size(0)
data = data.cuda()
target = target.cuda()
weights = alpha(data.size(0))
logits = model(data, weights)
loss = criterion(logits, target)
# Calculate the accuracy.
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
objs.update(loss.data, n)
top1.update(prec1.data, n)
top5.update(prec5.data, n)
if step % args.report_freq == 0 or step == len(valid_queue) - 1:
objs_avg = utils.reduce_tensor(objs.avg, args.world_size)
top1_avg = utils.reduce_tensor(top1.avg, args.world_size)
top5_avg = utils.reduce_tensor(top5.avg, args.world_size)
logging.info('valid %03d %e %f %f', step, objs_avg, top1_avg,
top5_avg)
return top1_avg, objs_avg