process raw data to graph, labels and masks
(self)
| 86 | ) |
| 87 | |
| 88 | def process(self): |
| 89 | """process raw data to graph, labels and masks""" |
| 90 | coo_adj = sp.load_npz(os.path.join(self.raw_path, "adj_full.npz")) |
| 91 | g = from_scipy(coo_adj) |
| 92 | |
| 93 | features = np.load(os.path.join(self.raw_path, "feats.npy")) |
| 94 | features = F.tensor(features, dtype=F.float32) |
| 95 | |
| 96 | y = [-1] * features.shape[0] |
| 97 | with open(os.path.join(self.raw_path, "class_map.json")) as f: |
| 98 | class_map = json.load(f) |
| 99 | for key, item in class_map.items(): |
| 100 | y[int(key)] = item |
| 101 | labels = F.tensor(np.array(y), dtype=F.int64) |
| 102 | |
| 103 | with open(os.path.join(self.raw_path, "role.json")) as f: |
| 104 | role = json.load(f) |
| 105 | |
| 106 | train_mask = np.zeros(features.shape[0], dtype=bool) |
| 107 | train_mask[role["tr"]] = True |
| 108 | |
| 109 | val_mask = np.zeros(features.shape[0], dtype=bool) |
| 110 | val_mask[role["va"]] = True |
| 111 | |
| 112 | test_mask = np.zeros(features.shape[0], dtype=bool) |
| 113 | test_mask[role["te"]] = True |
| 114 | |
| 115 | g.ndata["feat"] = features |
| 116 | g.ndata["label"] = labels |
| 117 | g.ndata["train_mask"] = generate_mask_tensor(train_mask) |
| 118 | g.ndata["val_mask"] = generate_mask_tensor(val_mask) |
| 119 | g.ndata["test_mask"] = generate_mask_tensor(test_mask) |
| 120 | |
| 121 | if self._reorder: |
| 122 | self._graph = reorder_graph( |
| 123 | g, |
| 124 | node_permute_algo="rcmk", |
| 125 | edge_permute_algo="dst", |
| 126 | store_ids=False, |
| 127 | ) |
| 128 | else: |
| 129 | self._graph = g |
| 130 | |
| 131 | def has_cache(self): |
| 132 | graph_path = os.path.join(self.save_path, "dgl_graph.bin") |
nothing calls this directly
no test coverage detected