SHREC2016dataset の型について

このサイトの実装を行っています。
training_jobs/mesh_classification/shrec_16/training_job_1.pyを実行すると、
pd_mesh_net/utils/datasets.pyの201行目

primal_graph_xs = torch.empty([0, dataset[0][0].x.shape[1]])

で

NotImplementedError

というエラーが表示されます。

試しに、datasetを表示させてみたところ、

Shrec2016DualPrimal(480, categories=['alien', 'ants', 'armadillo', 'bird1', 'bird2', 'camel', 'cat', 'centaur', 'dino_ske', 'dinosaur', 'dog1', 'dog2', 'flamingo', 'glasses', 'gorilla', 'hand', 'horse', 'lamp', 'laptop', 'man', 'myScissor', 'octopus', 'pliers', 'rabbit', 'santa', 'shark', 'snake', 'spiders', 'two_balls', 'woman'])

と返ってきました。

しかし、dataset[0]　や　dataset[0][0]　をprintで表示させると、

NotImplementedError

というエラーが表示されます。これはどいうことなのでしょうか？

ここで使用しているshrec16というデータは、
このコードと同様に書かれていることが調べていて分かったのですが、このdatasetは文字列で返されているのでしょうか？

そうだとすると、

dataset[0][0].x.shape[1]]

これはどのように取得すればよいのでしょうか？

以下長いですが、コードになります。

datasets.py
1def create_dataset(dataset_name,
2                   compute_node_feature_stats=True,
3                   node_feature_stats_filename=None,
4                   **dataset_params):
5  
6    if (dataset_name == 'shrec_16'):
7        dataset = Shrec2016DualPrimal(**dataset_params)
8    elif (dataset_name == 'cubes'):
9        dataset = CubesDualPrimal(**dataset_params)
10    elif (dataset_name == 'coseg'):
11        dataset = CosegDualPrimal(**dataset_params)
12    elif (dataset_name == 'human_seg'):
13        dataset = HumanSegDualPrimal(**dataset_params)
14    elif(dataset_name == 'vessel'):
15        dataset = []
16    else:
17        raise KeyError(
18            f"No known dataset can be generated with the name '{dataset_name}'."
19        )
20
21    node_statistics = None
22    print(dataset)
23    if (compute_node_feature_stats):
24        dataset_params = dataset.input_parameters
25        (primal_graph_mean, primal_graph_std, dual_graph_mean,
26         dual_graph_std) = compute_mean_and_std(
27             dataset=dataset,
28             dataset_params=dataset_params,
29             filename=node_feature_stats_filename)
30        node_statistics = (primal_graph_mean, primal_graph_std, dual_graph_mean,
31                           dual_graph_std)
32    return dataset, node_statistics
33
34
35def compute_mean_and_std(dataset=None, dataset_params=None, filename=None):
36    
37    if (dataset_params is not None):
38        for param_keyword in ['mean', 'std']:
39            for graph_keyword in ['primal', 'dual']:
40                keyword = f"{graph_keyword}_{param_keyword}"
41                if (keyword in dataset_params):
42                    raise KeyError(
43                        f"The parameters of the input dataset already contain "
44                        f"an entry '{keyword}'. Exiting.")
45    file_exists = False
46    if (filename is not None):
47        # Load the data from disk, if the file exists.
48        if (os.path.exists(filename)):
49            file_exists = True
50    if (file_exists):
51        assert (dataset_params is not None)
52        assert (isinstance(dataset_params, dict))
53        try:
54            with open(filename, "rb") as f:
55                data_from_disk = pkl.load(f)
56        except IOError:
57            raise IOError(f"Error loading cache mean-std file '{filename}'. "
58                          "Exiting.")
59        # Check that the file contains the mean and standard deviation.
60        for keyword in ['primal', 'dual']:
61            if (f'{keyword}_mean' not in data_from_disk):
62                raise KeyError(
63                    f"Cached file '{filename}' does not contain the mean of "
64                    f"the {keyword}-graph node features. Exiting.")
65            if (f'{keyword}_std' not in data_from_disk):
66                raise KeyError(
67                    f"Cached file '{filename}' does not contain the standard "
68                    f"deviation of the {keyword}-graph node features. Exiting.")
69        # Check that the size of the dataset is compatible.
70        try:
71            size_dataset_of_file = data_from_disk['dataset_size']
72        except KeyError:
73            raise KeyError(
74                f"Cached file '{filename}' does not contain the dataset size. "
75                f"Exiting.")
76        current_dataset_size = len(dataset)
77        if (size_dataset_of_file != current_dataset_size):
78            warnings.warn("Please note that the current dataset has size "
79                          f"{current_dataset_size}, whereas the cached file ("
80                          f"'{filename}') was generated from a dataset of size "
81                          f"{size_dataset_of_file}.")
82
83        # Check that the parameters match.
84        for param_name, param_value in dataset_params.items():
85            if (param_name not in data_from_disk):
86                raise KeyError(
87                    f"Could not find dataset parameter {param_name} in the "
88                    f"cached file '{filename}'. Please provide a different "
89                    "filename.")
90            else:
91                if (data_from_disk[param_name] != param_value):
92                    raise ValueError(
93                        f"Cached file '{filename}' is incompatible with "
94                        f"current dataset. Expected parameter {param_name} to "
95                        f"be {param_value}, found "
96                        f"{data_from_disk[param_name]}. Please provide a "
97                        "different filename.")
98        for cached_param_name in dataset_params.keys():
99            if (cached_param_name in [
100                    'primal_mean', 'primal_std', 'dual_mean', 'dual_std'
101            ]):
102                continue
103            if (cached_param_name not in dataset_params):
104                raise KeyError(
105                    f"Cached file '{filename}' is incompatible with "
106                    "current dataset, as it contains parameter "
107                    f"{cached_param_name}, which is missing in the input "
108                    "dataset. Please provide a different filename.")
109        # Return the cached data.
110        primal_graph_mean = data_from_disk['primal_mean']
111        primal_graph_std = data_from_disk['primal_std']
112        dual_graph_mean = data_from_disk['dual_mean']
113        dual_graph_std = data_from_disk['dual_std']
114    else:
115        # Compute the mean and standard deviation of the node features from
116        # scratch.
117        primal_graph_xs = torch.empty([0, dataset[0][0].x.shape[1]])
118        print('len',primal_graph_xs.size())
119        dual_graph_xs = torch.empty([0, dataset[0][1].x.shape[1]])
120        for sample_idx, (primal_graph, dual_graph, _, _) in enumerate(dataset):
121            primal_graph_xs = torch.cat([primal_graph_xs, primal_graph.x])
122            dual_graph_xs = torch.cat([dual_graph_xs, dual_graph.x])
123        assert (len(dataset) == sample_idx + 1)
124        primal_graph_mean = primal_graph_xs.mean(axis=0).numpy()
125        primal_graph_std = primal_graph_xs.std(axis=0).numpy()
126        dual_graph_mean = dual_graph_xs.mean(axis=0).numpy()
127        dual_graph_std = dual_graph_xs.std(axis=0).numpy()
128        assert (np.all(
129            primal_graph_std > 10 * np.finfo(primal_graph_std.dtype).eps))
130        assert (np.all(
131            dual_graph_std > 10 * np.finfo(dual_graph_std.dtype).eps))
132
133        if (filename is not None):
134            # Save the values to file, together with the dataset parameters and
135            # the dataset size, if required.
136            if (dataset_params is None):
137                dataset_params = {}
138            output_values = {
139                **dataset_params, 'primal_mean': primal_graph_mean,
140                'primal_std': primal_graph_std,
141                'dual_mean': dual_graph_mean,
142                'dual_std': dual_graph_std,
143                'dataset_size': sample_idx + 1
144            }
145            try:
146                with open(filename, 'wb') as f:
147                    pkl.dump(output_values, f)
148            except IOError:
149                raise IOError(
150                    "Unable to save mean-std data to file at location "
151                    f"{filename}.")
152
153    return (primal_graph_mean, primal_graph_std, dual_graph_mean,
154            dual_graph_std)
155