search

HaozheQi / P2B

P2B: Point-to-Box Network for 3D Object Tracking in Point Clouds
189 stars 36 forks source link

Some questions about your netR_36.model #11

Open StiphyJay opened 4 years ago

StiphyJay commented 4 years ago

Hello, thanks for you job. I want to know that weather the model(netR_36.model) is trained in two GTX1080Ti gpu? if you are, when i want to test this model in a machine with one GTX1080Ti, this model could weather or not work? When i test it, there are some error in my terminal. I have modified the code in test_tracking.py when i test. parser.add_argument('--ngpu', type=int, default=1, help='# GPUs') os.environ["CUDA_VISIBLE_DEVICES"] = '0' the error is: Traceback (most recent call last): File "test_tracking.py", line 177, in netR.load_state_dict(torch.load(os.path.join(args.save_root_dir, args.model)))
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 845, in load_state_dict self.class.name, "\n\t".join(error_msgs))) RuntimeError: Error(s) in loading state_dict for Pointnet_Tracking:

I would appreciate it if you could reply. Thank you!

AmingWu commented 4 years ago

@StiphyJay , Do you download the label_02 dataset? Whether the dataset only contains the training set.

StiphyJay commented 4 years ago

@StiphyJay , Do you download the label_02 dataset? Whether the dataset only contains the training set.

Yes,I download the label_02 dataset, and my dataset directory only contains testing set(0019,0020). Is it possible that the problem with the model caused the operation error?

HaozheQi commented 4 years ago

Hi, @StiphyJay, yes, the netR_36.model is trained with two gpus to speed up the training phase. But I don't know how to run the model trained with two gpus on a single gpu in Pytorch. In practice, the model trained with two gpus will be tested using two gpus as well. This is a suboptimal solution. If you can figure out this problem, I would appreciate for you sharing.

StiphyJay commented 4 years ago

Hi, @StiphyJay, yes, the netR_36.model is trained with two gpus to speed up the training phase. But I don't know how to run the model trained with two gpus on a single gpu in Pytorch. In practice, the model trained with two gpus will be tested using two gpus as well. This is a suboptimal solution. If you can figure out this problem, I would appreciate for you sharing.

Thanks for your reply. Here is my new question? Could you help me solve it? Traceback (most recent call last): File "test_tracking.py", line 178, in netR.load_state_dict(torch.load(os.path.join(args.save_root_dir, args.model)))
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 845, in load_state_dict self.class.name, "\n\t".join(error_msgs))) RuntimeError: Error(s) in loading state_dict for DataParallel: size mismatch for module.vote_layer.0.conv.weight: copying a param with shape torch.Size([256, 259, 1]) from checkpoint, the shape in current model is torch.Size([256, 256, 1]).

HaozheQi commented 4 years ago

I think you might change the original code here.