train pspnet error

[lfdeep]

When I train pspnet, set batch=2, it appears：
[epoch 1], [iter 10 / 4249], [train main loss 1.44544], [train aux loss 2.11281]. [lr 0.0049986500]
[epoch 1], [iter 20 / 4249], [train main loss 1.35266], [train aux loss 1.68135]. [lr 0.0049971499]
Traceback (most recent call last):
File "./train/voc-psp_net/train.py", line 258, in
main()
File "./train/voc-psp_net/train.py", line 111, in main
train(train_loader, net, criterion, optimizer, curr_epoch, args, val_loader, visualize)
File "./train/voc-psp_net/train.py", line 119, in train
for i, data in enumerate(train_loader):
File "/share/home/math8/.conda/envs/pytorch0.4.0_py3.5/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 286, in next
return self._process_next_batch(batch)
File "/share/home/math8/.conda/envs/pytorch0.4.0_py3.5/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 307, in _process_next_batch
raise batch.exc_type(batch.exc_msg)
RuntimeError: Traceback (most recent call last):
File "/share/home/math8/.conda/envs/pytorch0.4.0_py3.5/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 57, in _worker_loop
samples = collate_fn([dataset[i] for i in batch_indices])
File "/share/home/math8/.conda/envs/pytorch0.4.0_py3.5/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 138, in default_collate
return [default_collate(samples) for samples in transposed]
File "/share/home/math8/.conda/envs/pytorch0.4.0_py3.5/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 138, in
return [default_collate(samples) for samples in transposed]
File "/share/home/math8/.conda/envs/pytorch0.4.0_py3.5/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 115, in default_collate
return torch.stack(batch, 0, out=out)
RuntimeError: invalid argument 0: Sizes of tensors must match except in dimension 0. Got 4 and 2 in dimension 1 at /opt/conda/conda-bld/pytorch_1524580978845/work/aten/src/TH/generic/THTensorMath.c:3586

i don't know how to solve it?

[wxy920801]

I meet this problem too

[lfdeep]

I meet this problem too

I have solved this problem.You should set train_batch_size>=2

[wxy920801]

I meet this problem too

I have solved this problem.You should set train_batch_size>=2

Thank you. I tried to set train_batch_size=4 or 6, while this didn't work. Maybe anything else to correct?

[lfdeep]

I meet this problem too

I have solved this problem.You should set train_batch_size>=2

Thank you. I tried to set train_batch_size=4 or 6, while this didn't work. Maybe anything else to correct?

You should fix the input large image to the same size.
args = {
'train_batch_size': 4,
'lr': 1e-2 / sqrt(16 / 4),
'lr_decay': 0.9,
'max_iter': 3e4,
#'longer_size': 512,
#'crop_size': 473,
'input_size': 473,
'stride_rate': 2 / 3.,
'weight_decay': 1e-4,
'momentum': 0.9,
'snapshot': '',
'print_freq': 10,
'val_save_to_img_file': True,
'val_img_sample_rate': 0.01, # randomly sample some validation results to display,
'val_img_display_size': 384,
}
train_joint_transform = joint_transforms.Compose([
#joint_transforms.Scale(args['longer_size']),
joint_transforms.Scale(args['input_size']),
joint_transforms.RandomRotate(10),
joint_transforms.RandomHorizontallyFlip()
sliding_crop = joint_transforms.SlidingCrop(args['input_size'], args['stride_rate'], voc.ignore_label)
This is my modified code, I fixed my input image to 473

[wxy920801]

wow, done!
Thanks!