A bug about processing SHREC2016 datasets

🐛 Bug

I met the bug of processing SHREC2016 dataset. I can generate the ref_ .ptfiles, but I cannot generate train_ .pt and test_ .pt files

To Reproduce

Steps to reproduce the behavior:

1. my python code

import torch_geometric.datasets as gdatasets
s = gdatasets.SHREC2016('./SHREC2016', 'Cuts', 'David')

1. and the error is (after downloading and extracting the files. i.e. these 2 steps works fine)

---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
<ipython-input-31-15833c950a50> in <module>
----> 1 s = gdatasets.SHREC2016('./SHREC2016', 'Cuts', 'David')

~/anaconda3/lib/python3.7/site-packages/torch_geometric/datasets/shrec2016.py in __init__(self, root, partiality, category, train, transform, pre_transform, pre_filter)
     72         self.cat = category.lower()
     73         super(SHREC2016, self).__init__(root, transform, pre_transform,
---> 74                                         pre_filter)
     75         self.__ref__ = torch.load(self.processed_paths[0])
     76         path = self.processed_paths[1] if train else self.processed_paths[2]

~/anaconda3/lib/python3.7/site-packages/torch_geometric/data/in_memory_dataset.py in __init__(self, root, transform, pre_transform, pre_filter)
     54                  pre_filter=None):
     55         super(InMemoryDataset, self).__init__(root, transform, pre_transform,
---> 56                                               pre_filter)
     57         self.data, self.slices = None, None
     58 

~/anaconda3/lib/python3.7/site-packages/torch_geometric/data/dataset.py in __init__(self, root, transform, pre_transform, pre_filter)
     81 
     82         self._download()
---> 83         self._process()
     84 
     85     @property

~/anaconda3/lib/python3.7/site-packages/torch_geometric/data/dataset.py in _process(self)
    125 
    126         makedirs(self.processed_dir)
--> 127         self.process()
    128 
    129         print('Done!')

~/anaconda3/lib/python3.7/site-packages/torch_geometric/datasets/shrec2016.py in process(self)
    143 
    144         torch.save(ref_data, self.processed_paths[0])
--> 145         torch.save(self.collate(train_list), self.processed_paths[1])
    146         torch.save(self.collate(test_list), self.processed_paths[2])
    147 

~/anaconda3/lib/python3.7/site-packages/torch_geometric/data/in_memory_dataset.py in collate(self, data_list)
    122         r"""Collates a python list of data objects to the internal storage
    123         format of :class:`torch_geometric.data.InMemoryDataset`."""
--> 124         keys = data_list[0].keys
    125         data = data_list[0].__class__()
    126 

IndexError: list index out of range

Expected behavior

Environment

• OS: Ubuntu 16.04
• Python version: 3.7.3 (Anaconda)
• PyTorch version: 1.1.0
• CUDA/cuDNN version: NO (CPU)
• GCC version: 5.4.0
• Any other relevant information: NO

Additional context

I use torch.load to load the generated refs_ .pt files. And loading them is all-right.

[rusty1s]

I should never commit and push to master on a smartphone 😉

[rusty1s]

Fixed the SHREC16 bug in master 😃