def distill_model(distillation_box, data_loader, optimizer, log_freq, device,
epoch):
metric_logger = misc_util.MetricLogger(delimiter=' ')
metric_logger.add_meter(
'lr', misc_util.SmoothedValue(window_size=1, fmt='{value:.6f}'))
header = 'Epoch: [{}]'.format(epoch)
lr_scheduler = None
if epoch == 0:
warmup_factor = 1.0 / 1000.0
warmup_iters = min(1000, len(data_loader) - 1)
lr_scheduler = main_util.warmup_lr_scheduler(optimizer, warmup_iters,
warmup_factor)
for images, targets in metric_logger.log_every(data_loader, log_freq,
header):
images = list(image.to(device) for image in images)
targets = [{k: v.to(device) for k, v in t.items()} for t in targets]
loss = distillation_box(images, targets)
optimizer.zero_grad()
loss.backward()
optimizer.step()
if lr_scheduler is not None:
lr_scheduler.step()
metric_logger.update(loss=loss)
metric_logger.update(lr=optimizer.param_groups[0]['lr'])
def train_model(model, optimizer, data_loader, device, epoch, log_freq):
model.train()
metric_logger = misc_util.MetricLogger(delimiter=' ')
metric_logger.add_meter('lr', misc_util.SmoothedValue(window_size=1, fmt='{value:.6f}'))
header = 'Epoch: [{}]'.format(epoch)
lr_scheduler = None
if epoch == 0:
warmup_factor = 1.0 / 1000.0
warmup_iters = min(1000, len(data_loader) - 1)
lr_scheduler = main_util.warmup_lr_scheduler(optimizer, warmup_iters, warmup_factor)
for images, targets in metric_logger.log_every(data_loader, log_freq, header):
images = list(image.to(device) for image in images)
targets = [{k: v.to(device) for k, v in t.items()} for t in targets]
loss_dict = model(images, targets)
losses = sum(loss for loss in loss_dict.values())
# reduce losses over all GPUs for logging purposes
loss_dict_reduced = misc_util.reduce_dict(loss_dict)
losses_reduced = sum(loss for loss in loss_dict_reduced.values())
loss_value = losses_reduced.item()
if not math.isfinite(loss_value):
print('Loss is {}, stopping training'.format(loss_value))
print(loss_dict_reduced)
sys.exit(1)
optimizer.zero_grad()
losses.backward()
optimizer.step()
if lr_scheduler is not None:
lr_scheduler.step()
metric_logger.update(loss=losses_reduced, **loss_dict_reduced)
metric_logger.update(lr=optimizer.param_groups[0]['lr'])
def evaluate(model, data_loader, device):
n_threads = torch.get_num_threads()
# FIXME remove this and make paste_masks_in_image run on the GPU
torch.set_num_threads(1)
cpu_device = torch.device('cpu')
model.eval()
metric_logger = misc_util.MetricLogger(delimiter=' ')
header = 'Test:'
coco = get_coco_api_from_dataset(data_loader.dataset)
iou_types = get_iou_types(model)
coco_evaluator = CocoEvaluator(coco, iou_types)
for id, (image, targets) in enumerate(
metric_logger.log_every(data_loader, 100, header)):
image = list(img.to(device) for img in image)
targets = [{k: v.to(device) for k, v in t.items()} for t in targets]
torch.cuda.synchronize()
model_time = time.time()
outputs = model(image)
outputs = [{k: v.to(cpu_device)
for k, v in t.items()} for t in outputs]
model_time = time.time() - model_time
# print(targets, outputs)
res = {
target['image_id'].item(): output
for target, output in zip(targets, outputs)
}
evaluator_time = time.time()
coco_evaluator.update(res)
evaluator_time = time.time() - evaluator_time
metric_logger.update(model_time=model_time,
evaluator_time=evaluator_time)
# exit()
#
# if id > 30:
# break
# gather the stats from all processes
metric_logger.synchronize_between_processes()
# print('Averaged stats:', metric_logger)
coco_evaluator.synchronize_between_processes()
# accumulate predictions from all images
coco_evaluator.accumulate()
coco_evaluator.summarize()
torch.set_num_threads(n_threads)
return coco_evaluator