for the following code:
class_weights = xview_class_weights_hard_mining(range(60)).to(device)
class_weights2 = xview_class_weights_hard_mining(range(60)).to(device2)
for epoch in range(opt.epochs):
epoch += start_epoch
# img_size = random.choice([19, 20, 21, 22, 23, 24, 25]) * 32
# dataloader = ListDataset(train_path, batch_size=opt.batch_size, img_size=img_size, targets_path=targets_path)
# print('Running image size %g' % img_size)
# Update scheduler
# if epoch % 25 == 0:
# scheduler.last_epoch = -1 # for cosine annealing, restart every 25 epochs
# scheduler.step()
# if epoch <= 100:
# for g in optimizer.param_groups:
# g['lr'] = 0.0005 * (0.992 ** epoch) # 1/10 th every 250 epochs
# g['lr'] = 0.001 * (0.9773 ** epoch) # 1/10 th every 100 epochs
# g['lr'] = 0.0005 * (0.955 ** epoch) # 1/10 th every 50 epochs
# g['lr'] = 0.0005 * (0.926 ** epoch) # 1/10 th every 30 epochs
ui = -1
rloss = defaultdict(float) # running loss
metrics = torch.zeros(4, 60)
for i, (imgs, targets) in enumerate(dataloader):
n = 4 # number of pictures at a time
for j in range(int(len(imgs) / n)):
targets_j = targets[j * n:j * n + n]
nGT = sum([len(x) for x in targets_j])
if nGT < 1:
continue
# print(torch.cuda.memory_allocated())
loss = model(imgs[j * n:j * n + n].to(device), targets_j, requestPrecision=True,
weight=class_weights, epoch=epoch)
loss2 = model2(imgs[j * n:j * n + n].to(device2), targets_j, requestPrecision=True,
weight=class_weights2, epoch=epoch)
I am getting the following error:
Traceback (most recent call last):
File "traingpu.py", line 234, in <module>
main(opt)
File "traingpu.py", line 148, in main
weight=class_weights2, epoch=epoch)
File "/sdcc/u/amalik/.local/lib/python3.5/site-packages/torch/nn/modules/module.py", line 477, in __call__
result = self.forward(*input, **kwargs)
File "/gpfshome01/u/amalik/OGA/Yolov3/xview-yolov3/models.py", line 234, in forward
x, *losses = module[0](x, targets, requestPrecision, weight, epoch)
File "/sdcc/u/amalik/.local/lib/python3.5/site-packages/torch/nn/modules/module.py", line 477, in __call__
result = self.forward(*input, **kwargs)
File "/gpfshome01/u/amalik/OGA/Yolov3/xview-yolov3/models.py", line 128, in forward
width = ((w.data * 2) ** 2) * self.anchor_w
RuntimeError: arguments are located on different GPUs at /pytorch/aten/src/THC/generated/../generic/THCTensorMathPointwise.cu:314
I have two GPUs. I want to run two independent models on these two GPUs.