Memory loading error - Githubissues

Hi, @jprothero

I was testing your newest tweaking but I found an error about cpu vs gpu:

python main.py 

/home/hangyu5/anaconda2/envs/mctsnet/lib/python3.6/site-packages/torch/serialization.py:367: SourceChangeWarning: source code of class 'MCTSnet_model.MCTSnet' has changed. you can retrieve the original source code by accessing the object's source attribute or set `torch.nn.Module.dump_patches = True` and use the patch tool to revert the changes.
  warnings.warn(msg, SourceChangeWarning)
Loaded best model
Loaded best model
Loading memories...
Number of memories: 169
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 10/10 [00:23<00:00,  2.37s/it]
Saving memories...
HBox(children=(IntProgress(value=0, description='Epoch', max=1), HTML(value='')))
  0%|                                                                                                                                                                                | 0/10 [00:00<?, ?it/s]
Traceback (most recent call last):
  File "main.py", line 32, in <module>
    trainer.fastai_train(mctsnet.new, memories)
  File "/home/hangyu5/Desktop/Deep_Learning/MCTSnet/MCTSnet_trainer.py", line 80, in fastai_train
    learner.fit(5e-3, epochs, wds=1e-7) #was 7e-2
  File "/home/hangyu5/anaconda2/envs/mctsnet/lib/python3.6/site-packages/fastai/learner.py", line 99, in fit
    self.fit_gen(self.model, self.data, layer_opt, n_cycle, **kwargs)
  File "/home/hangyu5/anaconda2/envs/mctsnet/lib/python3.6/site-packages/fastai/learner.py", line 89, in fit_gen
    metrics=metrics, callbacks=callbacks, reg_fn=self.reg_fn, clip=self.clip, **kwargs)
  File "/home/hangyu5/anaconda2/envs/mctsnet/lib/python3.6/site-packages/fastai/model.py", line 84, in fit
    loss = stepper.step(V(x),V(y))
  File "/home/hangyu5/anaconda2/envs/mctsnet/lib/python3.6/site-packages/fastai/model.py", line 43, in step
    loss = raw_loss = self.crit(output, y)
  File "/home/hangyu5/Desktop/Deep_Learning/MCTSnet/MCTSnet_trainer.py", line 24, in train_wrapper
    return self.train(self.net, self.memories)
  File "/home/hangyu5/Desktop/Deep_Learning/MCTSnet/MCTSnet_trainer.py", line 40, in train
    states = torch.cat(states, dim=0)
RuntimeError: Expected a Tensor of type torch.cuda.FloatTensor but found a type torch.FloatTensor for sequence element 2  in sequence argument at position #1 'tensors'

The main error is

RuntimeError: Expected a Tensor of type torch.cuda.FloatTensor but found a type torch.FloatTensor for sequence element 2  in sequence argument at position #1 'tensors'

I guess the model you committed was trained on cpu alone, right? And when I load the model onto gpu, it expect different type of variable? Do you know any solution to fix this kind of issue in pytorch? (Delete your model and memory.* solve this problem, but I am still confused about cpu vs gpu in pytorch a bit)

jprothero / MCTSnet

Memory loading error #2