Hello My GPU is 1080T. However, There has the problem of insufficient video memory

GPU Insufficient memory！！ about pytorch-yolov3 HOT 3 CLOSED

eriklindernoren commented on July 27, 2024

GPU Insufficient memory！！

from pytorch-yolov3.

Comments (3)

pengzhiwang commented on July 27, 2024 1

Whether the code should be changed to the following form？

batch_size = int(hyperparams['batch'])
subdivisions = int(hyperparams['subdivisions'])
sub_batch = batch_size//subdivisions

.................

for epoch in range(opt.epochs):
for batch_i, (_, imgs, targets) in enumerate(dataloader):
imgs = Variable(imgs.type(Tensor))
targets = Variable(targets.type(Tensor), requires_grad=False)

    optimizer.zero_grad()
    for i in range(subdivisions):
        img = imgs[i*sub_batch:(i+1)*sub_batch]
        target = targets[i*sub_batch:(i+1)*sub_batch]
        loss = model(img, target)
        loss.backward()
        print('[Epoch %d/%d, Batch %d/%d] [Losses: x %f, y %f, w %f, h %f, conf %f, cls %f, total %f, recall: %.5f]' %
            (epoch, opt.epochs, batch_i, len(dataloader),
            model.losses['x'], model.losses['y'], model.losses['w'],
            model.losses['h'], model.losses['conf'], model.losses['cls'],
            loss.item(), model.losses['recall']))

    optimizer.step()
    model.seen += imgs.size(0)

if epoch % opt.checkpoint_interval == 0:
    model.save_weights('%s/%d.weights' % (opt.checkpoint_dir, epoch))

from pytorch-yolov3.

mark-hoffmann commented on July 27, 2024 1

You don't need to create sub_batches, just simply lower your batch size and you should be fine. It essentially does the same thing and that's what that parameter is for.

from pytorch-yolov3.

pengzhiwang commented on July 27, 2024

Due to limited conditions， I can only use one GPU

from pytorch-yolov3.

Recommend Projects