Unable to test models

[lccatala]

Hi, if I try to test a model I run into one of these two issues:

1. If I try to test the model provided at the repository in datasets/CamVid/ckpt-enet.pth, independently of using or not CUDA, I get the following error message:

Traceback (most recent call last):
  File "init.py", line 153, in <module>
    test(FLAGS)
  File "C:\Users\User\Desktop\ENet-Real-Time-Semantic-Segmentation\test.py", line 32, in test
    out1 = enet(tmg.float()).squeeze(0)
  File "C:\Users\User\Anaconda2\envs\tfg_temp\lib\site-packages\torch\nn\modules\module.py", line 532, in __call__
    result = self.forward(*input, **kwargs)
  File "C:\Users\User\Desktop\ENet-Real-Time-Semantic-Segmentation\models\ENet.py", line 194, in forward
    x, i1 = self.b10(x)
  File "C:\Users\User\Anaconda2\envs\tfg_temp\lib\site-packages\torch\nn\modules\module.py", line 532, in __call__
    result = self.forward(*input, **kwargs)
  File "C:\Users\User\Desktop\ENet-Real-Time-Semantic-Segmentation\models\RDDNeck.py", line 110, in forward
    x_copy = torch.cat((x_copy, extras), dim = 1)
RuntimeError: Expected object of backend CUDA but got backend CPU for sequence element 0 in sequence argument at position #1 'tensors'

2. If I try to run the model I've trained, I get the following error message:

Traceback (most recent call last):
  File "init.py", line 153, in <module>
    test(FLAGS)
  File "C:\Users\User\Desktop\ENet-Real-Time-Semantic-Segmentation\test.py", line 24, in test
    enet.load_state_dict(checkpoint['state_dict'])
  File "C:\Users\User\Anaconda2\envs\tfg_temp\lib\site-packages\torch\nn\modules\module.py", line 829, in load_state_dict
    raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
RuntimeError: Error(s) in loading state_dict for ENet:
        size mismatch for fullconv.weight: copying a param with shape torch.Size([16, 102, 3, 3]) from checkpoint, the shape in current model is torch.Size([16, 12, 3, 3]).

[AvivSham]

As for (1) can you share how you have loaded the weights to the model?

About (2) did you change anything in the model architecture?

[Riwaly]

As for(1), after loading the model, you can copy to cuda, just like this:
enet = ENet(12)
enet.load_state_dict(checkpoint['state_dict'])
enet = enet.to(FLAGS.cuda)
and then, copy image data to cuda,
tmg_ = plt.imread(FLAGS.image_path)
tmg_ = cv2.resize(tmg_, (h, w), cv2.INTER_NEAREST)
tmg = torch.tensor(tmg_).unsqueeze(0).float()
tmg = tmg.transpose(2, 3).transpose(1, 2)
tmg = tmg.to(FLAGS.cuda)

[qinghew]

@AvivSham i have met the (1) question. And i just use the pretrain model to do test.

[qinghew]

@Riwaly @AvivSham I get the answer, the --cuda in code is not work, you need load model, weights and picture into gpu, and it will go.